From matt@advanced.org  Tue Aug  3 17:17:44 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA14757
	for <ippm-archive@odin.ietf.org>; Tue, 3 Aug 1999 17:17:44 -0400 (EDT)
Received: from galatea (localhost [127.0.0.1])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with SMTP id RAA11133;
	Tue, 3 Aug 1999 17:17:46 -0400 (EDT)
From: "Matthew J Zekauskas" <matt@advanced.org>
To: <ippm-archive@ietf.org>
Cc: <ippm-request@advanced.org>
Subject: subscribe
Date: Tue, 3 Aug 1999 17:27:14 -0400
Message-ID: <000901beddf6$f44af900$4467dec7@galatea.advanced.org>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3155.0
Importance: Normal
Content-Transfer-Encoding: 7bit

Testing official IETF archive address, which has been added to the
IPPM list.

--Matt



From matt@advanced.org  Wed Aug  4 13:17:16 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA16985
	for <ippm-archive@odin.ietf.org>; Wed, 4 Aug 1999 13:17:15 -0400 (EDT)
Received: from krdl.org.sg (rodin.krdl.org.sg [192.122.139.27])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id NAA16276
	for <ippm@advanced.org>; Wed, 4 Aug 1999 13:06:06 -0400 (EDT)
Received: from mailhost.krdl.org.sg (mailbox.krdl.org.sg [192.122.134.30])
	by krdl.org.sg (8.9.3/8.9.3) with ESMTP id BAA04135
	for <ippm@advanced.org>; Thu, 5 Aug 1999 01:12:30 +0800 (SGT)
Received: from spitfire.krdl.org.sg (spitfire [192.122.135.125])
	by mailhost.krdl.org.sg (8.9.3/8.9.3) with SMTP id BAA16310;
	Thu, 5 Aug 1999 01:04:48 +0800 (SGT)
Received: by spitfire.krdl.org.sg (SMI-8.6/SMI-SVR4)
	id BAA23203; Thu, 5 Aug 1999 01:02:10 +0800
From: mtam@krdl.org.sg (Ivan Tam Ming Chit)
Message-Id: <199908041702.BAA23203@spitfire.krdl.org.sg>
Subject: Traceroute using larger packets
To: ippm@advanced.org
Date: Thu, 5 Aug 1999 01:02:10 +0800 (SGT)
Cc: mtam@mailbox.krdl.org.sg (Ivan Tam Ming Chit)
X-Mailer: ELM [version 2.4 PL25]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit

Hi :

  Would like to find out is there any implementation
  or way to change traceroute on Linux so that it will
  support packet size larger than 1500 bytes, e.g., 
  2800 bytes ? It complains about packet size too large, 
  though I thought IP layer will take care of segmentation.

  The very same traceroute implementation when 
  compiled on solaris allows larger packet size.

  Wonder why ?

  Thanks.

  -Ivan

   


From matt@advanced.org  Wed Aug  4 13:17:23 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA17000
	for <ippm-archive@odin.ietf.org>; Wed, 4 Aug 1999 13:17:22 -0400 (EDT)
Received: from unknown (ats1-5.worldramp.net [207.30.147.7])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with SMTP id NAA15733;
	Wed, 4 Aug 1999 13:07:56 -0400 (EDT)
From: sales@canada.com
Subject: Animated and Interactive Greeting Cards - Win $500 Cash -AVD
Date: Wed, 4 Aug 1999 13:09:33
Message-Id: <497.754124.679805@unknown>
Reply-To: cardstakes@canada.com
Mime-Version: 1.0
Content-Type: text/html; charset="us-ascii"


<BASE HREF="http://wallstreetresearch.net/PWDR/report.html">

<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<HTML>
<HEAD>
   <META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
   <META NAME="Author" CONTENT="Bret VanHorne: Beach Studios, Multimedia Website Design and Construction, www.BeachStudios.com">
   <META NAME="GENERATOR" CONTENT="Mozilla/4.05 [en]C-PBI-NC404  (Win95; U) [Netscape]">
   <TITLE>Wall Street Research: Reports</TITLE>
</HEAD>
<BODY TEXT="#000000" BGCOLOR="#CCCCCC" LINK="#9999FF" VLINK="#9999FF" MARGINHEIGHT="0" MARGINWIDTH="0">
&nbsp;
                <TABLE BORDER=0 CELLSPACING=0 CELLPADDING=0 WIDTH="100%" >
                        <TR>
                                <TD>
                                        <TABLE BORDER=0 CELLSPACING=0 CELLPADDING=3 WIDTH="100%" >
                                        <TR>
                                        <TD BGCOLOR="#666699"><B><FONT SIZE=+1>&nbsp; WALL STREET RESEARCH GROUP</FONT></B></TD>

                                        <td BGCOLOR="#666699"><a href="http://WallStreetResearch.net/" TARGET="_top">HOME</a></td>
                                        </TR>
                                        </TABLE>
                                </TD>
                        </TR>

                                <TR>
                                        <TD><IMG SRC="http://wallstreetresearch.net/images/spacer.gif" HEIGHT=10 WIDTH=1></TD>
                                </TR>

                                <TR>
                                        <TD ALIGN=RIGHT>
                                        <TABLE BORDER=0 CELLSPACING=0 CELLPADDING=2 WIDTH="95%" >
                                <TR>
                                        <TD>Buy Recommendation</TD>

                                        <TD ALIGN=RIGHT>July 26, 1999</TD>
                                </TR>

<TR>
<TD COLSPAN="2" BGCOLOR="#666699"><IMG SRC="http://wallstreetresearch.net/images/spacer.gif" HEIGHT=1 WIDTH=1></TD>
</TR>

<TR>
<TD COLSPAN="2">
<CENTER><U><FONT FACE="Swis721 BlkEx BT"><FONT SIZE=+1>Research Analysis</FONT></FONT></U>&nbsp;</CENTER>

<CENTER><FONT SIZE=+1>POWER DIRECT, INC.</FONT>&nbsp;</CENTER>

<CENTER>(OTC BB: PWDR)</CENTER>
</TD>
</TR>

<TR>
<TD ALIGN=CENTER COLSPAN="2">
<TABLE BORDER CELLSPACING=0 CELLPADDING=3 BORDERCOLOR="#000000" >
<TR>
<TD VALIGN=TOP WIDTH="320"><FONT SIZE=-1>Shares Outstanding: 17,327,500</FONT>&nbsp;
<BR><FONT SIZE=-1>Active Float: 13,127,500</FONT>&nbsp;
<BR><FONT SIZE=-1>Recent Price: $ .47</FONT>&nbsp;
<BR><FONT SIZE=-1>52 Week Bid Range: $.23 - $1.48</FONT>&nbsp;
<BR><FONT SIZE=-1>Capitalization: $8,131,000</FONT>&nbsp;</TD>

<TD ALIGN=RIGHT VALIGN=TOP WIDTH="245">
<TABLE BORDER=0 CELLSPACING=0 CELLPADDING=0 >
<TR>
<TD><FONT SIZE=-1>1999 EPS: $ .31(Est)</FONT>&nbsp;
<BR><FONT SIZE=-1>2000 EPS: $1.39(Est)</FONT>&nbsp;
<BR><FONT SIZE=-1>P/E Ratio: 1.5/1(Est)</FONT>&nbsp;
<BR><FONT SIZE=-1>Debt: NONE</FONT>&nbsp;
<BR><FONT SIZE=-1>Dividend Yield: N/A</FONT>&nbsp;</TD>

</TR>
</TABLE>
&nbsp;</TD>
</TR>
</TABLE>
&nbsp;</TD>
</TR>

<TR>
<TD COLSPAN="2"><IMG SRC="http://wallstreetresearch.net/images/spacer.gif" HEIGHT=10 WIDTH=1></TD>
</TR>

<TR>
<TD ALIGN=CENTER COLSPAN="2">
<TABLE BORDER CELLSPACING=0 BORDERCOLOR="#000000" >
<TR>
<TD WIDTH="197"><FONT SIZE=-1>Year Ending December 31, 1999</FONT></TD>

<TD WIDTH="72"><FONT SIZE=-1>1999 (E)</FONT></TD>

<TD WIDTH="72"><FONT SIZE=-1>2000 (E)</FONT></TD>

<TD WIDTH="72"><FONT SIZE=-1>2001 (E)</FONT></TD>

<TD WIDTH="72"><FONT SIZE=-1>2002 (E)</FONT></TD>

<TD WIDTH="72"><FONT SIZE=-1>2003 (E)</FONT></TD>
</TR>

<TR>
<TD><FONT SIZE=-1>Revenues (millions)</FONT></TD>

<TD><FONT SIZE=-1>$28.2 Mil</FONT></TD>

<TD><FONT SIZE=-1>$109 Mil</FONT></TD>

<TD><FONT SIZE=-1>$161 Mil</FONT></TD>

<TD><FONT SIZE=-1>$236 Mil</FONT></TD>

<TD><FONT SIZE=-1>$347 Mil</FONT></TD>
</TR>

<TR>
<TD><FONT SIZE=-1>Earnings (millions)</FONT></TD>

<TD><FONT SIZE=-1>$5.4 Mil</FONT></TD>

<TD><FONT SIZE=-1>$27.2 Mil</FONT></TD>

<TD><FONT SIZE=-1>$41.8 Mil</FONT></TD>

<TD><FONT SIZE=-1>$63.7 Mil</FONT></TD>

<TD><FONT SIZE=-1>$97.1 Mil</FONT></TD>
</TR>

<TR>
<TD><FONT SIZE=-1>Shares Outstanding (millions)</FONT></TD>

<TD><FONT SIZE=-1>17.3 Mil</FONT></TD>

<TD><FONT SIZE=-1>19.5 Mil</FONT></TD>

<TD><FONT SIZE=-1>21 Mil</FONT></TD>

<TD><FONT SIZE=-1>22 Mil</FONT></TD>

<TD><FONT SIZE=-1>25 Mil</FONT></TD>
</TR>

<TR>
<TD><FONT SIZE=-1>Earnings per share</FONT></TD>

<TD><FONT SIZE=-1>$ .31</FONT></TD>

<TD><FONT SIZE=-1>$1.39</FONT></TD>

<TD><FONT SIZE=-1>$1.99</FONT></TD>

<TD><FONT SIZE=-1>$2.89</FONT></TD>

<TD><FONT SIZE=-1>$3.88</FONT></TD>
</TR>

<TR>
<TD><FONT SIZE=-1>Projected Price P/E Ratio of 20/1</FONT></TD>

<TD><FONT SIZE=-1>$6.20</FONT></TD>

<TD><FONT SIZE=-1>$27.80</FONT></TD>

<TD><FONT SIZE=-1>$39.80</FONT></TD>

<TD><FONT SIZE=-1>$57.80</FONT></TD>

<TD><FONT SIZE=-1>$77.60</FONT></TD>
</TR>
</TABLE>
&nbsp;</TD>
</TR>

<TR>
<TD COLSPAN="2">

<P><B><FONT SIZE=+1>Business Summary:</FONT></B>&nbsp;

<p><b>Power Direct, Inc. (OTC BB: PWDR)</b> is a Company investing in viable business opportunities globally, with particular emphasis on the Internet and the opportunities available in the multi-billion dollar e-commerce area. 
<p><b><i>Power Direct's latest acquisition is the world's first "Greeting Card Based" sweepstakes game</i>. Cardstakes.com</b> <u>(http://www.cardstakes.net)</u> consists of an electronic Greeting Card that is sent to the recipient via e-mail enabling the recipient to play an "Instant Win" ticket for cash and participate in sweepstakes for huge jackpot prizes. With the same ticket the recipient and the sender also participate in two draws, a monthly and a yearly draw. Both the monthly and the yearly draws are shared by both the sender and the recipient.
<p><b><u>Statistics indicate that this untapped business sector has a world-wide market of over U.S. $15 Billion per year</u></b>. <b>Cardstakes.com will be the leader in this lucrative industry by being the</b> <b><i>FIRST</i></b> <b>online greeting card sweepstakes on the Internet</b>. Half of all annual greeting card sales are seasonal with the remaining being birthday and everyday occasion cards. <i>Cardstakes.com cards will be of high quality featuring special effects, animation, music, and custom design abilities.</i> 
<p><b>GREETING CARDS, SWEEPSTAKES & THE INTERNET - A WINNING COMBINATION</b>
<p><b><i>Combining three of the world's most profitable growth industries into one viable and powerful marketing platform, and being the first to do so, virtually insures that Power Direct, Inc. will be immensely successful.</i></b>

<p>PWDR expects to capture the lions' share of this new business by being the first to combine two of the world's oldest and most popular traditions, Greeting Cards and Sweepstakes, with the explosive growth of the Internet. 

<p>The traditional greeting card industry is a tremendously large and thriving business! <b><u>In 1998, over 7 billion greeting cards were purchased by American consumers, generating a total of $7.5 billion in U.S. retail sales</u></b>.  Of the total greeting cards purchased annually, roughly half are seasonal and the remaining half are everyday cards. Sales of alternative cards, especially non-occasion cards, are also on the increase.

<p><b>The traditional Instant Lottery and 6/49 style ticket sales in the USA and Canada in 1998 were over $42 billion U.S. dollars.</b>

<p>According to the <i>Computer Industry Almanac</i>, there were more than <b>147 million Internet users in the world at the end of 1998</b>. 

<p><b>This source further projects that there will be 320 million Internet users worldwide at the end of the year 2000, and more than 720 million users worldwide at the end of 2005. The U.S. Internet population will grow to more than 207 million Internet users in 2005, which will represent approximately 29 percent of the world's total.</b> 

<p>Therefore, PWDR feels that it has created a niche market that will <b><i><u>far outperform</u></i></b> some of the more traditional Internet Lottery concepts - because it is able to capture the entire population since everyone sends Greeting Cards.
<p><b>MORE FACTS ON THE HUGE GREETING CARD MARKET</b>
<p><b>The most popular card-sending holidays are, in order, Christmas, Valentine's Day, Mother's Day, Easter and Father's Day</b>. People of all ages and types exchange greeting cards. Women purchase more than 80 percent of all greeting cards, and the average card purchaser is a woman in her middle years, although this historically steady demographic picture may be changing. Cards range in price from $0.38 to $10.00, with the average card retailing for around $2.00. Cards featuring special techniques and new technologies are at the top of this price scale. 

<p>The average person receives 24 cards per year, seven of which are birthday cards. Estimates indicate that there are more than 1,800 greeting card publishers in America ranging from major corporations to small family organizations. Greeting Card Association members together account for approximately 90 percent of the industry market share. <b><i>A typical family sends out 125 greeting cards per annum</i></b>.       
<br>&nbsp;
<table cellspacing=0 cellpadding=0 width=60%>   
<tr>
<td><b><i>Greeting Card Sales broken down by Occasion:</i></b></td>
</tr>
</table>

<table cellspacing=0 cellpadding=0 width=80%>
<tr>
<td colspan=2 align=left></td>
<td><b><i><u>Seasons</u></i></b></td>
<td><b><i><u>Units</u></i></b></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Birthdays</i></td>
<td><i>2.7 billion</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Christmas</i></td>
<td><i>2.6 billion</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Valentine's Day</i></td>
<td><i>900 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Mother's Day</i></td>
<td><i>150 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Easter</i></td>
<td><i>120 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Father's Day</i></td>
<td><i>95 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Graduation</i></td>
<td><i>60 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Thanksgiving</i></td>
<td><i>30 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Halloween</i></td>
<td><i>25 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>St. Patrick's Day</i></td>
<td><i>15 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Jewish New Year</i></td>
<td><i>10 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Chanukah</i></td>
<td><i>10 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>New Year's</i></td>
<td><i>8 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>National Grandparent's Day</i></td>
<td><i>3 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Sweetest Day</i></td>
<td><i>1.5 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Passover</i></td>
<td><i>1.5 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Professional Secretary's Day</i></td>
<td><i>1 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>National Boss' Day</i></td>
<td><i>1 million</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>April Fool's Day</i></td>
<td><i>500,000</i></td>
</tr>
<tr>
<td colspan=2 align=left></td>
<td><i>Nurses' Day</i></td>
<td><i>500,000</i></td>
</tr>
</table>

<br>&nbsp;
<table cellspacing=0 cellpadding=0 width=60%>   
<tr>
<td align=center><b><i>Total U.S. Retail Sales</i></b></td>
</tr>
</table>
<br>&nbsp;



<table cellspacing=0 cellpadding=0 width=80%>
<tr>
<td colspan=4 align=left></td>
                <td>1998</td>
                <td>$7.5 bllion</td>
                <td>1988</td>
<td>$7.5 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1997</td>
                <td>$7.3 bllion</td>
                <td>1987</td>
<td>$3.8 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1996</td>
                <td>$6.8 bllion</td>
                <td>1986</td>
<td>$3.7 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1995</td>
                <td>$6.3 bllion</td>
                <td>1985</td>
<td>$3.5 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1994</td>
                <td>$5.9 bllion</td>
                <td>1984</td>
<td>$3.2 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1993</td>
                <td>$5.6 bllion</td>
                <td>1983</td>
<td>$2.7 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1992</td>
                <td>$5.3 bllion</td>
                <td>1982</td>
<td>$2.5bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1991</td>
                <td>$5.0 bllion</td>
                <td>1981</td>
<td>$2.35 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1990</td>
                <td>$4.6 bllion</td>
                <td>1980</td>
<td>$2.1 bllion</td>
</tr>
<tr>
<td colspan=4 align=left></td>
                <td>1989</td>
                <td>$4.2 bllion</td>
</tr>
</table>


<p>Information from the Greeting Card Association web site (http://www.greetingcard.org/gca/facts.htm )

<p>In addition to the regular holiday cards, PWDR will provide senders with a large assortment of non-occasional cards for use at any time of the year.

<p><b>MARKETING STRATEGY WILL PRODUCE EXPLOSIVE GROWTH</b>

<p>Being the first on the market and utilizing innovative marketing techniques, PWDR expects to capture a significant percentage of this huge niche market. Power Direct plans on utilizing some of the most unique Internet Direct Marketing plans and systems available today.

<p><b><i>Most of the sales will come from "word of mouth" style of marketing (most effective), but by e-mail and not by voice, which sets the stage for exponential growth for PWDR</i></b>.

<p>The initial marketing plan is to e-mail out to millions of users, a <b>FREE</b> Greeting Card with an imbedded Sweepstakes ticket with the potential to win $5,000.00 (or more).

<p>The Instant Win Portion of the Sweepstakes ticket is also a winner for every player.  But instead of winning money, each player wins a gift certificate good for 10 free promotional Cardstakes.com greeting cards to send out to 10 of his or her friends and family.

<p>Each of these promotional Cardstakes.com cards is of the same nature as the one he or she received directly from Power Direct, Inc.

<p><b><u>Based upon this scenario, it is anticipated that PWDR will have  millions of clients in a very short period of time - as one person sends to 10 people, and those 10 send to 100 people, and those 100 send to 1,000 people and so on.  The numbers build at an extremely rapid pace</u></b>.

<p>Multiply this by the number of users who will receive the initial e-mailed cards and the sky is the limit!

<p>Additional marketing plans include e-mailing free coupons, invitations, and other marketing materials to the millions of users of the cardstakes.com system on a regular basis.

<p><b>HOW IT WORKS  -  DYNAMICS & FUNDAMENTALS</b>

<p><b>Instead of spending $3 to $6 on a card that ends up at a land fill in a week, one can send an animated, singing, speaking, personally customized, virtual card through the Internet for just $3.00</b>. The card contains a scratch section that offers instant winnings of up to $1,000 and  other  prizes. The ticket number on the card will be entered into the sweepstakes jackpot of the month, (jackpot could be $5,000 and up), and the year-end grand jackpot (possibly $100,000) which is drawn on a yearly basis. Both the recipient and sender will share the prize money.

<p>Should you want to review the cards in the future, they will remain in your Cardstakes.com mailbox until you delete them.  There is no time limit like other Internet greeting cards.

<p><b>REVENUE ALLOCATION</b>

<p>As previously stated, the cards are sold for $3.00.  Thirty-five percent of the purchase price is kept by Power Direct, Inc. as a royalty.

<p><b>ECOLOGICAL BENEFIT</b>

<p><i>Thousands and thousands of tons of cards printed with toxic colored ink are dumped on the landfill each year</i>. <b>The "Virtual Sweepstakes" card will help to eliminate this huge unnecessary waste and save our planet</b>.  The cards not only provide a high-quality, sophisticated greeting card, but they also allow the sender a high level of interaction in the designing and viewing process.  First, there are 500 ready to send stock cards, (from art deco, vogue, classic, Victorian, and cartoons to 3D animation), thousands of available clip art and pictures, and speak as you type audio capabilities.   (For example the card will say "<i>Hi!  John, thank you for a wonderful time, love, Susan White</i>".  If you are willing to pay for the premium card, you could have your own voice and pictures on the card.  

<p><b><u>This is a "state of the art" product, not to be confused with the typical low quality free greeting card available on the Internet</u></b>.


<p><b>A TRUE "WIN-WIN" LEADING TO IMMEASURABLE GROWTH POTENTIAL</b>

<p><b>The growth potential of this business is immeasurable</b>. Imagine if one lady sends 100 Christmas cards to her friends, and if each of her friends sends out 100 and so on.  That is 10,000 cards generated by just one customer. This effect will, in fact, keep growing year after year. The customers of Power Direct become their unpaid sales persons, advertising their cards to every corner of the world.

<p>Let say, John won the jackpot of the month. He will receive $5,000 and Susan, the sender will receive $2,000.  The more cards Susan sends, the more chances she will win. There will be a <b>preferred customer bonus</b>. <i>Anyone who has sent out over 100 cards in a calendar year will receive a free bonus sweepstakes for every 50 cards</i>.  This once-a-year jackpot will be 100% awarded to a lucky sender.  All winners will be contacted through e-mail or their mailing address. The money will be deposited in an offshore bankers trust account ready for their money transfer instruction.


<p><b>INCREDIBLY TALENTED & SUCCESSFUL MANAGEMENT TEAM</b>

<p><b>Michael Wright (Chairman of the Board of Directors)</b> 
<p>Mr. Michael Wright has valuable executive experience in the gaming and lottery industries.  Mr. Wright has held such positions as Director of Distribution for Bally Gaming Inc., an affiliate of Las Vegas based Bally Casinos.  Bally Gaming Inc. is a licensed manufacturer of legal slot machines and video lottery terminals that are distributed internationally.  In 1997 he was President of San Diego based On-point Technology Systems, Inc. (ONTP: NASDAQ), which specializes in lottery ticket and high-tech point of sale vending machines.  Mr. Wright has first hand expertise in the establishment of new lotteries in Central Europe after he resided in Prague, Czech Republic to oversee the successful start-up of the nationally telecast TV BINGO NOVA lottery in 1994.
<p>Mr. Wright and Mr. Niessen worked together to set up the TV BINGO game in the Czech Republic.

<p><b>Brian Niessen (Game Design & Software Systems Controller-External Consultant)</b>

<p>Mr. Niessen has expertise in the design of software, Internet sites, gaming systems, television production and operations necessary to run a successful lottery program.

<p>He is a recognized front runner in the development of "<i>interactive touch-screen</i>" electronic gaming systems and video game development as early as 1982.and was responsible for in the software development and technical production of the highly rated game shows "<i>TalkAbout</i>", "<i>The Last Word</i>", "<i>5-4-3-2-RUN</i>", produced in Canada and England; and broadcast in Canada, the United States and England.

Mr. Niessen and his team are credited with the success of the "<i>Televizni Bingo</i>" game in the Czech Republic during September 1994 though April 1996, during which time approximately US$ 44 million was paid out in prize winnings out of approximately U.S.$ 80 million gross revenue. He is currently involved with registered charities setting up fundraising sweepstakes in North America and abroad.

<dir>
<b>Employment History</b>
<p>NET Software Inc., President.
        <br><spacer size=30>Video Game Software Company
        <br><spacer size=30>International Video Game Development and Manufacturing Company.

<p>PirateBusters, Inc., President.
        <br><spacer size=30>Video Game Anti-Piracy Development Company
        <br><spacer size=30>Supplier to Electronic Arts, Inc., the largest Video Game manufacturer.

<p>DesignTech Business Systems, Inc., President.
        <br><spacer size=30>Provided all computer technology to popular television shows.
<br>FOX Network -<spacer size=72>"TalkAbout" - Number 3 rated TV Game Show in USA, 
<br>CBC Network -<spacer size=72>"TalkAbout" - Number 1 rated TV Game Show in Canada.
<br>FOX Network -<spacer size=72>"The Last Word" - Number 4 rated TV Game Show in USA.
<br>CTV Network -<spacer size=72>"The Last Word" - Number 2 rated TV Game Show in Canada.
<br>ITV Network, Leeds, UK - "TalkAbout"
<br>CTV Network -<spacer size=74>"5-4-3-2-RUN" - Children's Television Show.
<br>CTV Network -<spacer size=74>"Lingo"
<br>NBC Network -<spacer size=74>"Lingo"
<br>BC Place Stadium - Provided hardware and specialized software to perform animations on 20 x 40 foot screen.

<p>BNG Design Group, Inc.,President.
<br>Developer of Home Banking Systems for major Canadian Bank.

<p>DesignTech Electronic Gaming Systems, Inc.,President
<br>Developer and Manufacturer of Touch Screen Electronic Bingo/Gaming Systems.

<p>Skywalker Management, Inc., President.
<br>Developer of Software and Systems, Startup and Operations Manager of  Czech Republic TV Bingo Lottery.

<p>BingoNet Technologies, Ltd., President
            <br>Developer and Data Centre Operations of Lottery Management and Operational Systems.
        </dir>

<p><b>Conrado Beckerman</b>

<p>Conrado Beckerman has the qualifications to direct the international marketing and Internet Affiliation program.  

<p>Mr. Beckerman's assistance with his proven track record of developing international Telecommunications markets, studies in International Relations, cultural background in social, political & economic issues as well as his industrial sales experience and negotiation skills, intuitive understanding of international business dynamics, unique language skills, and understanding of Government Liaisons, Protocol, Diplomatic and International Business upbringing will greatly assist PWDR in expanding into the worldwide markets.

<p>Mr. Beckerman's accomplishments include: Developing International markets for various high technology companies, dealing successfully in transfer of technology agreements in Argentina and Brazil, developing national standards for wireless cable and Government of Uruguay. In addition, Mr. Beckerman consulted for wireless cable operators in Nepal, Pakistan and Indonesia, South Africa and Latin America.

<p>Mr. Beckerman's past positions include:

<p>1998 - present:
<br>Strategic Planning Director, Southern Cone
<br>Vitacom Corporation, Mountain View, Ca.

<p>Developed the market in the Southern Cone of South America for VSAT equipment bringing new accounts to the company. Marketed satellite space segment for PTT's.  Marketed equipment to ISP's in the region.  Introduced new Internet and Voice over Satellite equipment generating interest and achieving letters of intent and contracts worth thirty million dollars.  Participated  in seminars and conferences as a guest speaker. Developed strategy for local manufacturing in Brasil.

<p>1993 - 1998
<br>Honorary Consul for Uruguay
<br>Vancouver, Canada

<p>Achieved the highest position available as a foreign dignitary for the Provinces of British Columbia and Alberta. Promoted cultural and economic exchange between Uruguay and the Western Provinces. Speaker in various forums related to Mercosur. Promoted Uruguay as a pole for distribution to the Southern Cone. Involved in developing joint venture agreements related to technology investment in Uruguay.

<p>1996 - 1998 
<br>Marketing Director, Latin America
<br>Norsat International, Vancouver, B.C. 

<p>Designed Market strategy for Cable television encoding/set-top systems in each country of the region achieving sales of CAN$ 20 million in eleven months of work. Worked in joint venture agreement with Brazilian companies for local manufacturing. In addition, developed a Latin American long run strategy for the company to be involved in value added generating incomes from PPV. 

<p>1991-1996
<br>President, CB&A

<br>Vancouver, Canada.

<p>Distributor of satellite related equipment worldwide. Designed and installed first cable television network  in Brasil. Marketed MMDS equipment in South America. Made own OEM satellite receivers and TVRO equipment. Exhibited and attended in most important trade shows in the industry. Started and finalized joint venture agreement to manufacture low noise equipment in Brasil for M.T.I. Wrote "Televisão via Satellite" in collaboration with Dr. Frank Baylin. Writing for "America Economia" magazine telecommunications column. Participated in trade commissions invited by the Canadian Government. Represented various television transmitter manufactures and won important international bids. Spoke at various seminars and consulted investors on television opportunities in Latin America. Obtained license for MMDS station in Rancagua, Chile; later sold it to TCI.

<p>1988-1991
<br>International Sales Manager
<br>Norsat International, Surrey, B.C.

<p>Developed markets for consumer and commercial satellite receiving systems in South East Asia, Latin America, Middle East, Europe and Africa. Developed technology transfer agreement in Latin America for local manufacturing. Brought a whole range of receiving and transmission equipment for resale. Introduced products to new investors. 

<p>1985-1988
<br>International Sales Manager, Europe & Latin America
<br>Nexus Engineering Corp., Vancouver,  B.C.

<p>Managed Latin America and European market development and sales.

<p>1982-1986
<br>Radio Canada International, CBC, Vancouver, B.C.
<br>Reporter- Broadcaster
<br>Reporting and broadcasting economic news of British Colombia to Latin America

<p>1979
<br>UNESCO, Paris, France
<br>Assistant to the Ambassador

<p><b>SUMMATION:</b> 

<p><b>Power Direct, Inc. has a well thought out business plan with dynamic, experienced and very capable management, and is in the forefront of what will be a huge growth industry.  We believe PWDR is well positioned to reap a substantial share of the virtual greeting card industry.  Their unique niche, with the continuing and tremendous growth & interest of the greeting card industry and Internet combined with the Sweepstakes should pave the way for substantial capital appreciation in the shares of PWDR in the months and years to come</b>.

<p><b>RECOMMENDATION:  STRONG  BUY</b>

<p><b><i><u>BASED ON PWDR'S PROJECTED HIGH GROWTH RATE AND THEIR LEADERSHIP POSITION IN THIS DYNAMIC & HIGHLY PROFITABLE NICHE INDUSTRY, WE HAVE A VERY STRONG BUY RECOMMENDATION FOR BOTH NEAR AND LONG TERM APPRECIATION. WE FORECAST THAT THESE SHARES WILL BE TRADING AT $6.00 TO $8.00 BY JANUARY, 2000 AND $25.00 TO $35.00 BY JANUARY, 2001</u></i></b>.

<ul>
<p><b><li>Power Direct, Inc. is positioned to become the dominant leader in the multi-billion dollar "virtual" lottery greeting card industry</li></b>.
<p><b><li>Company's niche is in three huge, growing, & profitable areas - Lottery, Internet and Greeting Cards.  Their business strategy should easily generate hundreds of millions of dollars of revenues in the next few years with excellent profitability.</li></b>
<p><b><li>Management team is strong and focused & will contribute significantly to future success.  PWDR's business plan is excellent!</li></b>
<p><b><li>With high projected cash flow, Company will eventually be an obvious acquisition candidate.</li></b>
<p><b><li>Company will benefit substantially in any other e-commerce endeavors due to projected millions of clients from their virtual greeting card business.</li></b>
<p><b><li>Company's charitable contributions and ecological benefit will be viewed positively by many people, leading to increasing business.</li></b>
<p><b><li>The profit margins are very high in this business.  Earnings per share will be increasing substantially in a relatively short period of time.</li></b>
<p><b><li>Industry potential is projected at $15 Billion per year.  Capturing even a small percentage of that business assures PWDR of major profitability.

</li></b>
</ul>

<P>&nbsp;</TD>
</TR>

<TR>
<TD COLSPAN="2">
<TABLE BORDER CELLSPACING=0 CELLPADDING=20 BORDERCOLOR="#000000" >
<TR>
<TD><FONT SIZE=-2> 
This message complies with the PROPOSED United States Federal requirements for commercial email bill, Section 301,Paragraph 
(a)(2)(C) of S. 1618.

We Honor ALL Remove Requests!
Further sending of this e-mail can be stopped by visiting our site at http://www.cardstakes.net/wsr.html and following the prompts to remove your email address from our mailing list.


THIS IS A PUBLICATION OF WALL STREET RESEARCH GROUP,
A FINANCIAL COMMUNICATIONS AND INVESTOR RELATIONS COMPANY. WALL STREET
RESEARCH GROUP, A DIVISION OF WALL STREET FINANCIAL GROUP, INC., SERVES
AS SPECIAL ADVISOR TO THE FEATURED COMPANY AND HAS RECEIVED A BASE FEE
OF 75,000 SHARES OF FREE TRADING COMMON STOCK FOR 6 MONTHS REPRESENTATION
IN ADDITION TO EXPENSES FOR PREPERATION AND DISTRIBUTION OF MATERIALS INCLUDING
THIS REPORT. MORE DETAILED DISCLOSURE REGARDING FEES PAID IS AVAILABLE
UPON REQUEST. THIS IS NOT AN OFFER TO BUY OR SELL SECURITIES. INFORMATION
OR OPINIONS IN THIS REPORT ARE PRESENTED SOLELY FOR INFORMATIVE PURPOSES,
AND ARE NOT INTENDED NOR SHOULD THEY BE CONSTRUED AS INVESTMENT ADVICE.
COMPANIES MENTIONED HEREIN MAY CARRY A HIGH INVESTMENT RISK; READERS SHOULD
CAREFULLY REVIEW THE COMPANIES THOROUGHLY WITH THEIR REGISTERED INVESTMENT
ADVISOR OR REGISTERED STOCKBROKER. THE ANALYSIS CONTAINED HEREIN DOES NOT
PURPORT TO BE A COMPLETE STUDY OF THE FEATURED COMPANY OR OTHER COMPANIES
MENTIONED. INFORMATION USED AND STATEMENTS OF FACT HAVE BEEN OBTAINED FROM
THE FEATURED COMPANY AND OTHER SOURCES, BUT NOT VERIFIED NOR GUARANTEED
BY WALL STREET RESEARCH GROUP AS TO COMPLETENESS OR ACCURACY. SUCH INFORMATION
IS SUBJECT TO CHANGE WITHOUT NOTICE. OPINIONS STATED HEREIN MAY BE SOLELY
WALL STREET RESEARCH GROUP'S OR THE INDICATED SOURCES, AND NOT NECESSARILY
THOSE OF THE FEATURED COMPANY. YOU SHOULD ASSUME THAT OFFICERS, DIRECTORS,
AND EMPLOYEES OF WALL STREET RESEARCH GROUP OR OTC COMMUNICATIONS AND FINANCIAL
ANALYSTS MENTIONED, AND MEMBERS OF THEIR FAMILIES HOLD A POSITION AND MAY
FROM TIME TO TIME TRADE IN THESE SECURITIES FOR THEIR OWN ACCOUNTS. SPECIFIC
INFORMATION IN THIS REGARD WILL BE FURNISHED UPON REQUEST. TRADEMARKS ARE
THE PROPERTY OF THEIR RESPECTIVE OWNERS. &copy; 1998 WALL STREET RESEARCH
GROUP. ALL RIGHTS RESERVED. DUPLICATION OF THIS REPORT WITHOUT THE EXPRESSED
WRITTEN PERMISSION OF WALL STREET RESEARCH GROUP IS STRICTLY PROHIBITED.
FOR FREE REPRINTS OR TO BE PLACED ON WALL STREET RESEARCH GROUP'S MAILING
LIST, CALL: IN THE U.S.: 877-660-1932 OR OUTSIDE OF THE U.S., CALL: 714-664-1472.&nbsp;</FONT></TD>
</TR>
</TABLE>
&nbsp;</TD>
</TR>

<TR>
<TD COLSPAN="2"><IMG SRC="http://wallstreetresearch.net/images/spacer.gif" BORDER=0 HEIGHT=20 WIDTH=1></TD>
</TR>

<TR>
<TD COLSPAN="2" BGCOLOR="#666699"><IMG SRC="http://wallstreetresearch.net/images/spacer.gif" BORDER=0 HEIGHT=1 WIDTH=1></TD>
</TR>

<TR>
<TD ALIGN=CENTER COLSPAN="2">

<P><B>| <A HREF="http://wallstreetresearch.net/home.html">Home</A> | <A HREF="http://wallstreetresearch.net/research.html">Research</A> | <A HREF="http://wallstreetresearch.net/contact.html">Contact
Info</A> |</B>&nbsp;
<BR><B>| <A HREF="http://www.WallStreetResearch.net/" TARGET="_top">Frames</A>
| <A HREF="http://www.wallstreetresearch.net/PWDR/report.html" TARGET="_top">No-Frames</A>
|</B>&nbsp;
<BR>&nbsp;
<DIV ALIGN=right><FONT SIZE=-2>Copyright &copy; 1998 &amp; 1999 Wall Street
Research Group Inc. All Rights Reserved</FONT></DIV>
</TD>
</TR>
</TABLE>
&nbsp;</TD>
</TR>
</TABLE>
&nbsp;
</BODY>
</HTML>




From matt@advanced.org  Wed Aug  4 17:01:25 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA22761
	for <ippm-archive@odin.ietf.org>; Wed, 4 Aug 1999 17:01:24 -0400 (EDT)
Received: from galatea (localhost [127.0.0.1])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with SMTP id QAA28385
	for <ippm@advanced.org>; Wed, 4 Aug 1999 16:55:16 -0400 (EDT)
From: "Matthew J Zekauskas" <matt@advanced.org>
To: <ippm@advanced.org>
Subject: ADMIN: mailing list filter installed for UCE
Date: Wed, 4 Aug 1999 17:04:25 -0400
Message-ID: <004a01bedebc$eed66b00$4467dec7@galatea.advanced.org>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3155.0
Importance: Normal
Content-Transfer-Encoding: 7bit

The IPPM Mailing list has been hit twice in the past two
weeks with UCE ("unsolicited commercial email").  While
I'm reluctant to make changes based on one or two posts,
it appears as though we might have been added to some UCE
software's list.

As you all know, the intent of any IETF mailing list is to
remain completely open.  I therefore propose the following
change to stop most UCE, which I belive is consistent with
IETF practice.  Effective Real Soon Now, any mail
that is sent to the list which DOES NOT contain ippm@advanced.org
in the To: or Cc: fields will be bounced to me in my capacity
as list maintainer.  If the mail is at all on-topic I will forward
it to the list, otherwise toss it in the bit bucket.

The intent is to delay anything which is likely irrelevant, and
let everything else through immediately.

The bottom line: you probably don't have to do anything.  Just
make sure you have ippm@advanced.org in the to: or cc: of any mail
you intend to go to the list.

Matt Zekauskas
IP Performance Metrics Working Group Co-Chair & List Maintaner



From guest@advanced.org  Mon Aug  9 18:07:50 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA14312
	for <ippm-archive@odin.ietf.org>; Mon, 9 Aug 1999 18:07:49 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id RAA05907
	for ippm-l@advanced.org; Mon, 9 Aug 1999 17:57:24 -0400 (EDT)
Received: from thumper.research.telcordia.com (thumper.research.telcordia.com [128.96.41.1])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id RAA09471;
	Mon, 9 Aug 1999 17:57:23 -0400 (EDT)
Received: from avast.research.telcordia.com (avast [192.4.15.92])
	by thumper.research.telcordia.com (8.9.1a/8.9.1) with ESMTP id RAA11774;
	Mon, 9 Aug 1999 17:56:19 -0400 (EDT)
Received: (from wel@localhost)
	by avast.research.telcordia.com (8.8.8/8.8.8) id RAA11891;
	Mon, 9 Aug 1999 17:56:19 -0400 (EDT)
Date: Mon, 9 Aug 1999 17:56:19 -0400 (EDT)
From: "Will E. Leland" <wel@research.telcordia.com>
Message-Id: <199908092156.RAA11891@avast.research.telcordia.com>
To: minutes@ietf.org
Subject: minutes for IPPM WG, 13 July 1999
Cc: ippm@advanced.org, matt@advanced.org

Here's the minutes from our hour at Oslo; the presentations are
available in PDF format by http and ftp from
    http://www.advanced.org/IPPM/Meetings/ietf45/
-- Will

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Tuesday, 13 July 1999 at 14:15-15:15
====================================

The meeting was chaired by the working group chairs, Will Leland
and Matt Zekauskas, and was reported by Paul Love.

AGENDA
1. Introduction & Status
2. IPDV
3. Loss pattern update
4. Futures and Milestones

IETF Home page: http://www.ietf.org/html.charters/ippm-charter.html
IPPM Home page: http://www.advanced.org/IPPM/

1. Introduction & Status, Matt Zekauskas (Advanced Network & Services)

Within the last two weeks a number of metrics were approved as
Proposed Standard.  Vern Paxson and Scott Bradner will write a
document that explains the meaning of a metric to be on "standards
track" (similar to the document for MIBs).  Connectivity (RFC 2498)
will be reclassified as a Proposed Standard, and the one-way delay,
loss, and round-trip delay drafts were all approved as Proposed
Standards.

Since the last meeting, the bulk transfer framework, IP delay
variation (jitter) and loss pattern drafts were all updated.
If anyone has comments on the bulk transfer framework, please
take them to the mailing list; it seems to the chairs that
the bulk transfer document is close to being done (at least
until we have experience with some individual bulk transfer
metrics).  The IPDV document has been picked up by Phil Chimento
(see below).  The loss patterns document has been updated,
but mostly cosmetically, and will be discussed today.

2 IPDV Update, Phil Chimento (University of Twente)

Since the last meeting, Phil Chimento has taken over the
IP Delay Variation draft after Carlo Demichelis switched
positions; Phil has released version 3 of the draft.
The major changes are that the "discontinuous samples"
definition was deleted from the draft, and the old discussion
of IPDV mean value was also deleted.  A new section was
added to clarify distributions of variation, and a new statistic
was added for jitter that follows Van Jacobson's definition
from DiffServ.  In addition, a new section on the relationship
with other standards (ITU-T) was added, as well as a short
section on security.  The next version will clean up the
discussion of clock skew and the effects of taking IPDV
measurements when the clocks are not synchronized.

Phil then went through a discussion on the components of delay
(see slide) to ensure that we all understood what delay variation
really means, as presented in this draft.  The specific emphasis was
on the relationship between pair-wise delay variation versus change
in delay from some absolute:  even where each pair-wise delay variation
is small, the total variation can be large.

There was also a discussion of the ITU's delay variation measures
for 1-point and 2-point delay variation of ATM cells, and the
non-normative appendix to the IP measurement specification
which discusses 2-point delay for IP.  The ITU's specification
is based on absolute variation from a reference, either a
particular value (e.g., the first) or the average of a set
of values.

Finally, Phil turned to questions for the working group --
essentially is the IPDV metric headed in the right direction?

In particular, what is the purpose of IPDV?
- Jitter Estimation?  If so, in the EE sense (deviation from
  absolute) or the Computer Science sense (relative differences;
  for example, Van's definition from the expedited forwarding
  specification, or min & max differentiation with distribution)?	
- Sizing Buffers?
- Traffic characterization?
- Estimation of queuing delay?

What would you do if you had reams and reams of numbers?

One ISP in the audience said that their need was for measuring real
time voice characteristics and capacity planning.  These kinds
of uses imply that summaries and trends over time are important.
Another audience member mentioned that delay variation metrics
can serve to determine how large a token bucket is required to
provide unimpeded access to a packet stream that is intended to
be constant (or nearly constant) rate. Will mentioned that such
use relates to so-called "R/S statistics" developed in hydrology
for sizing dams.

3. Loss pattern update
	Introduction by Matt Zekauskas,
	Update by Steven Glass (Nokia)

Matt's introduction gave the basics about loss patterns and why
they are interesting, and about the loss pattern metrics which build
on one-way loss streams. [See slides.]

Steve pointed out the changes to the document, which include
many minor changes.  Section 3.3 added a clarifying example,
but it may have a typo.  There is a new Loss-distance-stream
metric.

Finally, Steve led a discussion about potential controversial
items and whether the metrics are ready to be submitted to the
IESG.  It is felt that the draft needs polish, but it's complete
and pretty close to being ready to go.

A question from the audience asked about the effect of the number
of drops causing retransmits and timeouts that cause retransmits
(specifically, TCP).  This particular item is out of scope for
the group, except as part of a bulk transfer capacity metric.
Matt Z. noted that Matt Mathis has talked about a "TCP MIB" --
exporting TCP variables.  However, that is out-of-scope for us.

There was a discussion about "Noticeable Rate" -- some applications
are sensitive to even two adjacent packet losses.

There was also a discussion about the fact that a burst loss
is considered terminated by a single good packet arrival.
A burst followed by a single good packet and another burst
would seem to be better characterized as one long burst.
Will noted that what is being measured could provide the
raw data for constructing a "less sensitive" metric.
Would some sort of "errored period" be a meaningful thing
to define?

Finally, there was a discussion about what to do with this draft.
Is it useful enough to move forward?  Should it be standards track?
Consensus from the group was that it should be cleaned up
and moved forward as an Experimental RFC.

4. Milestones and futures, Matt Zekauskas and Will Leland (Telcordia).

There wasn't much time left, but Matt presented the current thinking
on milestones.  [See the slides.] Key to the table:  The columns are
target dates -- months in 1999 and 1Q 2000.  The work items are on the
left.  Within the table, N means new draft, R means revised draft,
S means submit to the IESG.  "IPPM and ITU" is shorthand for a draft
that compares the IPPM metrics with the ITU metrics based on Will
Leland's presentation at the last IETF in Minneapolis.  The Delay-Loss
BCP is a best-current-practices document for the one-way delay and
loss metrics, based on at least the experience with the Advanced
Network & Services implementation and the RIPE NCC implementation.
The milestones are fairly aggressive, but the goal is to have all
the basic metrics done by 1Q 2000.

There were two questions from the floor: should Internet weather
maps be discussed here, and should we define MIBs for metrics.  Both
topics are apparently in-scope, as long as the maps are restricted
to the specific metrics under discussion.  However, they are not
currently specifically part of the charter.  There has been a
preference in the IETF for MIBs to be developed by the working group
that originates the related standards, so MIBs for metrics naturally
fall under our responsibility. We should look for help from people
already experienced in MIB design and use.

In continued informal discussions after the official WG slot ended,
the broad question was raised about the applicability of IPPM's
proposed metrics for SLAs: are our current metrics of use for SLAs?
Are there other metrics within our scope that could be useful for
SLA definitions? The Chairs urged that this discussion be raised
on the mailing list.



From guest@advanced.org  Tue Aug 10 04:03:40 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id EAA04126
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 04:03:40 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id DAA07701
	for ippm-l@advanced.org; Tue, 10 Aug 1999 03:58:43 -0400 (EDT)
Received: from universe.serc.rmit.edu.au (universe.serc.rmit.edu.au [131.170.42.12])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id DAA17524
	for <ippm@advanced.org>; Tue, 10 Aug 1999 03:58:40 -0400 (EDT)
Received: (from darryl@localhost)
	by universe.serc.rmit.edu.au (8.8.5/8.8.5) id RAA26325
	for ippm@advanced.org; Tue, 10 Aug 1999 17:58:31 +1000 (EST)
Date: Tue, 10 Aug 1999 17:58:31 +1000 (EST)
From: Darryl Veitch <darryl@serc.rmit.edu.au>
Message-Id: <199908100758.RAA26325@universe.serc.rmit.edu.au>
To: ippm@advanced.org
Subject: Loss patterns,     Re: minutes for IPPM WG, 13 July 1999
X-Sun-Charset: US-ASCII

Hello all,

In a recent mail from Will Leland on the meeting in Oslo, the following comment 
was made:

&&&&&&&&&&&&&&&&&&&&&&&&
"  There was also a discussion about the fact that a burst loss
is considered terminated by a single good packet arrival.
A burst followed by a single good packet and another burst
would seem to be better characterized as one long burst.
Will noted that what is being measured could provide the
raw data for constructing a "less sensitive" metric.
Would some sort of "errored period" be a meaningful thing
to define?  "
&&&&&&&&&&&&&&&&&&&&&&&&

Two students and I did some work last year looking at characterising 
the loss process in detail.
One of the issues to clarify was precisely this one of, 'ok losses tend to occur
in small groups, but what about patterns such as  111011 ? 
The results we obtained were quite surprising. We found that the structure of losses
in UDP test packet streams (send out either as Poisson or with constant separation)
was extremely 'independent', in fact looking like an alternating renewal process
in most cases we observed (~100 experiments over 15 routes in 8 countries).

That is, take two random variables  L (Loss) and A (Arrival). 
 To a good approximation the loss process consisted in simply
taking sample values alternately from L and A, in an independent fashion.
This offers the prospect of a simple description of the loss process consisting
of a discrete probability distribution for L (with most of the mass on 1, 2 or 3
losses), and an exponential distribution for A.  
With this underlying model the probability of obtaining patterns 
such as 111011, or any others, is easily calculated.

I know it is late days as far as the RFC is concerned, but I think that such a 
description could be of use.   I will be taking up this work in more detail in
coming months.    Sue Moon presented similar finding at Infocom this year.

The paper appeared in Globecom 98 and can be downloaded from  
 http://www.serc.rmit.edu.au/~darryl     (paper at the bottom)

 

Darryl Veitch

+-------------------------------+------------------------------------------+
|      Dr. Darryl Veitch 	| Email:   darryl@serc.rmit.edu.au         |
|      Senior Research Fellow   |                                          |
|      S.E.R.C.                 | Telephone:                               |
|      110 Victoria St          |        Direct-   +61 3 9925 4014         |
|      Carlton 3053             |     Inquiries-   +61 3 9925 4013         |
|      GPO Box 2476V            |                                          |
|      Melbourne                | Fax:             +61 3 9925 4094         |
|      Victoria 3001            |                                          |
|      Australia                | Web: http://www.serc.rmit.edu.au/~darryl |
+-------------------------------+------------------------------------------+



From guest@advanced.org  Tue Aug 10 04:27:40 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id EAA04177
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 04:27:39 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id EAA11264
	for ippm-l@advanced.org; Tue, 10 Aug 1999 04:23:00 -0400 (EDT)
Received: from birch.ripe.net (birch.ripe.net [193.0.1.96])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id EAA26326;
	Tue, 10 Aug 1999 04:22:58 -0400 (EDT)
Received: from x49.ripe.net (x49.ripe.net [193.0.1.49])
	by birch.ripe.net (8.8.8/8.8.8) with ESMTP id KAA06687;
	Tue, 10 Aug 1999 10:22:24 +0200 (CEST)
Received: from localhost (henk@localhost)
	by x49.ripe.net (8.8.8/8.8.5) with SMTP id KAA18862;
	Tue, 10 Aug 1999 10:22:24 +0200 (CEST)
X-Authentication-Warning: x49.ripe.net: henk owned process doing -bs
Date: Tue, 10 Aug 1999 10:22:24 +0200 (CEST)
From: "Henk Uijterwaal (RIPE-NCC)" <henk@ripe.net>
To: "Will E. Leland" <wel@research.telcordia.com>
cc: ippm@advanced.org, Matthew J Zekauskas <matt@advanced.org>
Subject: Random comments (was Re: minutes for IPPM WG, 13 July 1999)
In-Reply-To: <199908092156.RAA11891@avast.research.telcordia.com>
Message-ID: <Pine.BSI.3.95.990810095248.17885H-100000@x49.ripe.net>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

(Editted the distribution list a bit).

On Mon, 9 Aug 1999, Will E. Leland wrote:

> Here's the minutes from our hour at Oslo; 

> 4. Milestones and futures, Matt Zekauskas and Will Leland (Telcordia).

I don't see it anywhere in the minutes and I'm not even sure anymore if I
made the comment in Oslo :-), but I think that we need a (BCP?) document
describing how to compare numbers and distributions.  Contrary to to
protocols and other binary stuff, this is NOT immediately obvious. 

For example:  Suppose two implementations measure a delay over the same
path for a time-interval and quote that as (10%, median, 90%).  Now
implementation 1 finds (15ms, 30ms, 45ms) and #2 finds (13ms, 26ms, 39ms),
does that mean that the values are comparable and that the difference is
due to statistical fluctuations and sampling errors _or_ is there a
significant difference between the two?

The same goes for distributions.  Look back at the slides I showed in
M'polis.  I think we really need some agreed method better than just
looking at them, to say that 2 distributions are compatible. 

> There were two questions from the floor: should Internet weather
> maps be discussed here, and should we define MIBs for metrics.  Both
> topics are apparently in-scope, as long as the maps are restricted
> to the specific metrics under discussion.  However, they are not
> currently specifically part of the charter.  

I believe the charter should be modified in this way.  During the last
months, I attended a number of meetings with the customers of ISP's that
are (or will) host one of our test-boxes, and the two things that always
came up were: 

1. Everybody seems interested in a "quality of the Internet" map, but 
   nobody seems to be able to define what should go in there (and if
   the IPPM metrics are indeed useful for this).  This group might be
   able to define what we can (not) provide for an Internet Weather Map

2. Can we use the results of your (delay/loss) measurements in our
   SLA.  People put values for delays/losses in SLA's but nobody 
   seems to put any method for measuring them in the SLA.  I think
   it would be useful if we could provide some "building blocks" 
   for SLA's.  Not the values, more something along the lines of

    - Metric
    - Method of measuring the metric
    - Numbers that are quoted
    - When is the result acceptable.  (This is related to the previous
      point, if I put "Median Delay = X ms", in an SLA is (X-1, X+0.1,
      X+100) as (10%, median, 50%) in or outside that limit?)

   That way, both vendor and customer know what is being measured,
   how it is measured and when the SLA isn't met.

Does anybody else think that this is useful?


And finally, during my vacation after Oslo, I read a paper on Pathchar and
started to wonder if we could measure the effective throughput of a link
using 1-way delays with different packetsizes.  Has anybody ever tried
this?

Henk

------------------------------------------------------------------------------
Henk Uijterwaal                    Email: henk.uijterwaal@ripe.net
RIPE Network Coordination Centre     WWW: http://www.ripe.net/home/henk
Singel 258                         Phone: +31.20.535-4414,  Fax -4445
1016 AB Amsterdam                   Home: +31.20.4195305
The Netherlands                   Mobile: +31.6.55861746  
------------------------------------------------------------------------------

The Committee (...) was unable to reach a consensus that substantial merit was
lacking. Thus, the appeal was deemed meritorious.          (Orlando NABC #19).



From guest@advanced.org  Tue Aug 10 09:39:44 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA12810
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 09:39:43 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id JAA11515
	for ippm-l@advanced.org; Tue, 10 Aug 1999 09:35:07 -0400 (EDT)
Received: from tadarida.theasis.com (tadarida.theasis.com [206.9.240.247])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id JAA12544
	for <ippm@advanced.org>; Tue, 10 Aug 1999 09:35:05 -0400 (EDT)
Received: from localhost (andy@localhost)
	by tadarida.theasis.com (8.9.2/8.9.2) with SMTP id IAA17546;
	Tue, 10 Aug 1999 08:34:55 -0500 (CDT)
X-Authentication-Warning: tadarida.theasis.com: andy owned process doing -bs
Date: Tue, 10 Aug 1999 08:34:54 -0500 (CDT)
From: Andy Scherrer <andy@mids.org>
X-Sender: andy@tadarida.theasis.com
To: Darryl Veitch <darryl@serc.rmit.edu.au>
cc: ippm@advanced.org
Subject: Re: Loss patterns,     Re: minutes for IPPM WG, 13 July 1999
In-Reply-To: <199908100758.RAA26325@universe.serc.rmit.edu.au>
Message-ID: <Pine.GSO.3.96.990810081313.16501C-100000@tadarida.theasis.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


> Two students and I did some work last year looking at characterising the
> loss process in detail.  One of the issues to clarify was precisely this
> one of, 'ok losses tend to occur in small groups, but what about
> patterns such as 111011 ?  The results we obtained were quite
> surprising. We found that the structure of losses in UDP test packet
> streams (send out either as Poisson or with constant separation)  was
> extremely 'independent', in fact looking like an alternating renewal
> process in most cases we observed (~100 experiments over 15 routes in 8
> countries). 
> 
> That is, take two random variables L (Loss) and A (Arrival). 
>  To a good approximation the loss process consisted in simply taking
> sample values alternately from L and A, in an independent fashion. 

But in what proportions?  Of course, that's a question you ask of the
data, but in a sequence whose beginning and end are difficult to define,
the time scale becomes an issue. And so while loss of one packet may not
be dependent on loss of any other packet, there is dependency on time
scale, since it will probably matter if you're looking at such proportions
in 3 sec bursts vs. 30 sec or 3 min bursts.

> This offers the prospect of a simple description of the loss process
> consisting of a discrete probability distribution for L (with most of
> the mass on 1, 2 or 3 losses), and an exponential distribution for A. 

But note that using a simple univariate exponential distribution for A
implies a Poisson distribution. This group has discussed problems with
Poisson modelling of these data. That is not to say it shouldn't perhaps
be discussed more, perhaps with some thought to heirarchical modeling of
the Poisson parameter (lambda) -- that would be one approach for
addressing the time issue I mentioned above. 

> With this underlying model the probability of obtaining patterns such as
> 111011, or any others, is easily calculated. 

Well, I submit that there are actually a suite of underlying models to
consider. A binary outcome such as your example can result from one of
many distributions, each of which thinks of the "question" in a different
way. 

For example, if you think of it as Binomial problem (a series of
Bernoulli[0-1] trials), you have a fixed, known number of trials (N), 
and model the proportion out of N that are 1 (or 0).

However, another way of looking at the _same_ data could be "How many
trials (packets) do I see before I observe a 1"? Doing this repeatedly
might give a set of data consisting of sequence lengths.

 ...which is different from "How many trials (packets) do I observe before
I can count some number C of 1s?" This can also give data consisting of
sequence lengths, but counted differently. 

> I know it is late days as far as the RFC is concerned, but I think that
> such a description could be of use.  I will be taking up this work in
> more detail in coming months.  Sue Moon presented similar finding at
> Infocom this year. 
> 
> The paper appeared in Globecom 98 and can be downloaded from  
>  http://www.serc.rmit.edu.au/~darryl     (paper at the bottom)

Thanks for the pointer - I'll take a look. 

Andy Scherrer
Statistician
Matrix Information and Directory Services, Inc.
1106 Clayton Lane, Ste. 501W, Austin, TX 78723
http://www.mids.org/
+1-512-451-7602


> 
> Darryl Veitch
> 
> +-------------------------------+------------------------------------------+
> |      Dr. Darryl Veitch 	| Email:   darryl@serc.rmit.edu.au         |
> |      Senior Research Fellow   |                                          |
> |      S.E.R.C.                 | Telephone:                               |
> |      110 Victoria St          |        Direct-   +61 3 9925 4014         |
> |      Carlton 3053             |     Inquiries-   +61 3 9925 4013         |
> |      GPO Box 2476V            |                                          |
> |      Melbourne                | Fax:             +61 3 9925 4094         |
> |      Victoria 3001            |                                          |
> |      Australia                | Web: http://www.serc.rmit.edu.au/~darryl |
> +-------------------------------+------------------------------------------+
> 
> 



From guest@advanced.org  Tue Aug 10 09:50:32 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA13829
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 09:50:31 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id JAA02137
	for ippm-l@advanced.org; Tue, 10 Aug 1999 09:47:47 -0400 (EDT)
Received: from tadarida.theasis.com (tadarida.theasis.com [206.9.240.247])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id JAA27371;
	Tue, 10 Aug 1999 09:47:45 -0400 (EDT)
Received: from localhost (andy@localhost)
	by tadarida.theasis.com (8.9.2/8.9.2) with SMTP id IAA17576;
	Tue, 10 Aug 1999 08:47:31 -0500 (CDT)
X-Authentication-Warning: tadarida.theasis.com: andy owned process doing -bs
Date: Tue, 10 Aug 1999 08:47:31 -0500 (CDT)
From: Andy Scherrer <andy@mids.org>
X-Sender: andy@tadarida.theasis.com
To: "Henk Uijterwaal (RIPE-NCC)" <henk@ripe.net>
cc: "Will E. Leland" <wel@research.telcordia.com>, ippm@advanced.org,
        Matthew J Zekauskas <matt@advanced.org>
Subject: Re: Random comments (was Re: minutes for IPPM WG, 13 July 1999)
In-Reply-To: <Pine.BSI.3.95.990810095248.17885H-100000@x49.ripe.net>
Message-ID: <Pine.GSO.3.96.990810083524.16501D-100000@tadarida.theasis.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

> > 4. Milestones and futures, Matt Zekauskas and Will Leland (Telcordia).
> 
> I don't see it anywhere in the minutes and I'm not even sure anymore if I
> made the comment in Oslo :-), but I think that we need a (BCP?) document
> describing how to compare numbers and distributions.  Contrary to to
> protocols and other binary stuff, this is NOT immediately obvious. 

Not immediately obvious, and also could get hairy. 
How to compare numbers depends on the numbers and on the distribution, not
to mention some measure of how the data were collected. 

But I agree that it makes a lot of sense to explore and document the realm
of cases that IPPM interests tend to encounter. 

> For example:  Suppose two implementations measure a delay over the same
> path for a time-interval and quote that as (10%, median, 90%).  Now
> implementation 1 finds (15ms, 30ms, 45ms) and #2 finds (13ms, 26ms, 39ms),
> does that mean that the values are comparable and that the difference is
> due to statistical fluctuations and sampling errors _or_ is there a
> significant difference between the two?

Adequate comparison needs even more information, including sample sizes.
All statistical comparisons are deemed "significant" or not on the
basis of variation around the points you want to compare (usually the
center). While the 10% and 90% values give a range that is useful for
quantifying variation, the weighting of such a quantity depends on sample
size. And also on how independent your samples are. 

And then, defining what qualifies as "significant" is an entirely
different problem, since at the very least there's a distinction between
"detectible" with a certain reliability and "important" as far as, say, a
user or application would be concerned. 

> The same goes for distributions.  Look back at the slides I showed in
> M'polis.  I think we really need some agreed method better than just
> looking at them, to say that 2 distributions are compatible. 

To come up with an improvement on just looking at them (noting that humans
are really good at pattern recognition), you'd have to identify features
of the distributions that are important. Once again it's a question of
weighting: how important are deviations in the tails, vs. deviations in
the center, or some other region?


[SLA-related questions.

> Does anybody else think that this is useful?

FWIW, I do. 
Looks like a lot happened in Oslo; sorry I missed it!

Andy Scherrer
Statistician
Matrix Information and Directory Services, Inc.
1106 Clayton Lane, Ste. 501W, Austin, TX 78723
http://www.mids.org/
+1-512-451-7602



> Henk
> 
> ------------------------------------------------------------------------------
> Henk Uijterwaal                    Email: henk.uijterwaal@ripe.net
> RIPE Network Coordination Centre     WWW: http://www.ripe.net/home/henk
> Singel 258                         Phone: +31.20.535-4414,  Fax -4445
> 1016 AB Amsterdam                   Home: +31.20.4195305
> The Netherlands                   Mobile: +31.6.55861746  
> ------------------------------------------------------------------------------
> 
> The Committee (...) was unable to reach a consensus that substantial merit was
> lacking. Thus, the appeal was deemed meritorious.          (Orlando NABC #19).
> 
> 



From guest@advanced.org  Tue Aug 10 10:52:26 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA17792
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 10:52:25 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id KAA20612
	for ippm-l@advanced.org; Tue, 10 Aug 1999 10:46:58 -0400 (EDT)
Received: from mw.3com.com (intergate.usr.com [149.112.20.3])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id KAA03784
	for <ippm@advanced.org>; Tue, 10 Aug 1999 10:46:57 -0400 (EDT)
Received: from mwgate02.mw.3com.com by mw.3com.com (8.8.5/3.1.090690-3Com Corporation)
	id JAA01688; Tue, 10 Aug 1999 09:51:20 -0500 (CDT)
Received: by mwgate02.mw.3com.com(Lotus SMTP MTA v4.6.4  (830.2 3-23-1999))  id 862567C9.0051F1B9 ; Tue, 10 Aug 1999 09:55:02 -0500
X-Lotus-FromDomain: 3COM@3COM-MWGATE
From: "Mike Borella" <Mike_Borella@mw.3com.com>
To: Darryl Veitch <darryl@serc.rmit.edu.au>
cc: ippm@advanced.org
Message-ID: <862567C9.0051EFE9.00@mwgate02.mw.3com.com>
Date: Tue, 10 Aug 1999 09:52:58 -0500
Subject: Re: Loss patterns, Re: minutes for IPPM WG, 13 July 1999
Mime-Version: 1.0
Content-type: text/plain; charset=us-ascii
Content-Disposition: inline



How do you mean 'independent'?  Uncorrelated?  Doesn't independent
mean that the arrival and loss burst processes would both be well modeled as
geometrically-distributed (at least in the Bernoulli sense)?

And how do the parameters vary over different times of day, days of week,
time scales and paths?  While the general model of a binary Markov chain
can be used to model this sort of process, if the parameters are all over the
map,
the model may not be very useful in practice.  With any luck the parameter
values can be ranged based on some physical observations.

I performed a simple analysis of UDP loss patterns a couple of years back on
6 months of data across 3 paths.  On all three paths, loss bursts fit a Pareto
better than anything else I tried.  The interesting thing was that bursts of 10
or
less packets lost always had to be modeled differently that bursts of about 15
or
greater.  Given the time scales involved (about 30 ms between packets) it seemed
as if the shorter loss bursts were caused by transient congestion while the
longer
bursts (up to and exceeding 1000's of packets) were cause by a more serious
outage such as downtime or a reboot.

-Mike





Darryl Veitch <darryl@serc.rmit.edu.au> on 08/10/99 02:58:31 AM

Sent by:  Darryl Veitch <darryl@serc.rmit.edu.au>


To:   ippm@advanced.org
cc:    (Mike Borella/MW/US/3Com)
Subject:  Loss patterns,     Re: minutes for IPPM WG, 13 July 1999




Hello all,

In a recent mail from Will Leland on the meeting in Oslo, the following comment
was made:

&&&&&&&&&&&&&&&&&&&&&&&&
"  There was also a discussion about the fact that a burst loss
is considered terminated by a single good packet arrival.
A burst followed by a single good packet and another burst
would seem to be better characterized as one long burst.
Will noted that what is being measured could provide the
raw data for constructing a "less sensitive" metric.
Would some sort of "errored period" be a meaningful thing
to define?  "
&&&&&&&&&&&&&&&&&&&&&&&&

Two students and I did some work last year looking at characterising
the loss process in detail.
One of the issues to clarify was precisely this one of, 'ok losses tend to occur
in small groups, but what about patterns such as  111011 ?
The results we obtained were quite surprising. We found that the structure of
losses
in UDP test packet streams (send out either as Poisson or with constant
separation)
was extremely 'independent', in fact looking like an alternating renewal process
in most cases we observed (~100 experiments over 15 routes in 8 countries).

That is, take two random variables  L (Loss) and A (Arrival).
 To a good approximation the loss process consisted in simply
taking sample values alternately from L and A, in an independent fashion.
This offers the prospect of a simple description of the loss process consisting
of a discrete probability distribution for L (with most of the mass on 1, 2 or 3
losses), and an exponential distribution for A.
With this underlying model the probability of obtaining patterns
such as 111011, or any others, is easily calculated.

I know it is late days as far as the RFC is concerned, but I think that such a
description could be of use.   I will be taking up this work in more detail in
coming months.    Sue Moon presented similar finding at Infocom this year.

The paper appeared in Globecom 98 and can be downloaded from
 http://www.serc.rmit.edu.au/~darryl     (paper at the bottom)



Darryl Veitch

+-------------------------------+------------------------------------------+
|      Dr. Darryl Veitch      | Email:   darryl@serc.rmit.edu.au         |
|      Senior Research Fellow   |                                          |
|      S.E.R.C.                 | Telephone:                               |
|      110 Victoria St          |        Direct-   +61 3 9925 4014         |
|      Carlton 3053             |     Inquiries-   +61 3 9925 4013         |
|      GPO Box 2476V            |                                          |
|      Melbourne                | Fax:             +61 3 9925 4094         |
|      Victoria 3001            |                                          |
|      Australia                | Web: http://www.serc.rmit.edu.au/~darryl |
+-------------------------------+------------------------------------------+








From guest@advanced.org  Tue Aug 10 10:59:04 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA17988
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 10:59:03 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id KAA08356
	for ippm-l@advanced.org; Tue, 10 Aug 1999 10:56:46 -0400 (EDT)
Received: from galatea (localhost [127.0.0.1])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with SMTP id KAA10106;
	Tue, 10 Aug 1999 10:56:42 -0400 (EDT)
From: "Matthew J Zekauskas" <matt@advanced.org>
To: "Phil Chimento" <chimento@cs.utwente.nl>
Cc: "Matt Zekauskas" <matt@advanced.org>, <ippm@advanced.org>
Subject: IPDV clock terminology
Date: Tue, 10 Aug 1999 11:06:43 -0400
Message-ID: <000001bee341$f51dc1a0$32f46680@galatea.advanced.org>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
Importance: Normal
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3155.0
Content-Transfer-Encoding: 7bit

I was re-reading the IPDV draft yesterday and found that one term
bothered me as I was reading the draft, and that was the reuse of
the Framework term "skew".

Carlo (I believe) coined the terms "skew" and "drift", where skew was
the fixed offset and drift was the variable offset of two clocks.

How about just calling the fixed offset "offset"?

--Matt



From guest@advanced.org  Tue Aug 10 11:04:08 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA18186
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 11:04:07 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id LAA08723
	for ippm-l@advanced.org; Tue, 10 Aug 1999 11:01:14 -0400 (EDT)
Received: from utrhcs.cs.utwente.nl (utrhcs.cs.utwente.nl [130.89.10.247])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id LAA26788;
	Tue, 10 Aug 1999 11:01:12 -0400 (EDT)
Received: from myrtilos.cs.utwente.nl (myrtilos.cs.utwente.nl [130.89.16.9])
	by utrhcs.cs.utwente.nl (8.9.1/8.9.1) with SMTP id RAA18514;
	Tue, 10 Aug 1999 17:01:09 +0200 (MET DST)
Received: from heisenberg.cs.utwente.nl by myrtilos.cs.utwente.nl (SMI-8.6/csrelay-Sol1.4/RB)
	id RAA25425; Tue, 10 Aug 1999 17:01:09 +0200
Received: from localhost by heisenberg.cs.utwente.nl (8.8.8+Sun/SMI-SVR4)
	id RAA10084; Tue, 10 Aug 1999 17:01:07 +0200 (MET DST)
Message-Id: <199908101501.RAA10084@heisenberg.cs.utwente.nl>
X-Mailer: exmh version 2.0.2 2/24/98
To: "Matthew J Zekauskas" <matt@advanced.org>
Cc: ippm@advanced.org
Subject: Re: IPDV clock terminology 
In-reply-to: Your message of "Tue, 10 Aug 1999 11:06:43 -0400."
             <000001bee341$f51dc1a0$32f46680@galatea.advanced.org> 
References: <000001bee341$f51dc1a0$32f46680@galatea.advanced.org> 
From: "Philip F. Chimento" <chimento@cs.utwente.nl>
X-Organisation: University of Twente, Dept. of Computer Science
                P.O. Box 217, NL-7500 AE Enschede, The Netherlands
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Tue, 10 Aug 1999 17:01:07 +0200
Sender: chimento@cs.utwente.nl

Hi Matt: 
Ok. That is fine with me. As I said in Oslo I need to go over those 
sections carefully and clarify them (and make sure that the 
terminology is in conformance with the framework). Any other 
suggestions like this are welcome. 
Regards, Phil 
-- 
-------------------------------------------------------------------------
       ___o___  Phil Chimento                       tel.: +31 53 4894331
   ___  | | |   <chimento@cs.utwente.nl>            fax : +31 53 4894524
  /     | | |   Centre for Telematics and Information Technology  (CTIT)
  \___  | | |   University of Twente, P.O. Box 217, 7500 AE Enschede, NL




From guest@advanced.org  Tue Aug 10 12:07:28 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA19696
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 12:07:27 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id MAA25804
	for ippm-l@advanced.org; Tue, 10 Aug 1999 12:01:07 -0400 (EDT)
Received: from tadarida.theasis.com (tadarida.theasis.com [206.9.240.247])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id MAA12740;
	Tue, 10 Aug 1999 12:01:05 -0400 (EDT)
Received: from localhost (andy@localhost)
	by tadarida.theasis.com (8.9.2/8.9.2) with SMTP id LAA17928;
	Tue, 10 Aug 1999 11:00:51 -0500 (CDT)
X-Authentication-Warning: tadarida.theasis.com: andy owned process doing -bs
Date: Tue, 10 Aug 1999 11:00:51 -0500 (CDT)
From: Andy Scherrer <andy@mids.org>
X-Sender: andy@tadarida.theasis.com
To: "Philip F. Chimento" <chimento@cs.utwente.nl>
cc: Matthew J Zekauskas <matt@advanced.org>, ippm@advanced.org
Subject: Re: IPDV clock terminology 
In-Reply-To: <199908101501.RAA10084@heisenberg.cs.utwente.nl>
Message-ID: <Pine.GSO.3.96.990810105107.16501O-100000@tadarida.theasis.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII



On Tue, 10 Aug 1999, Philip F. Chimento wrote:

> Hi Matt: 
> Ok. That is fine with me. As I said in Oslo I need to go over those 
> sections carefully and clarify them (and make sure that the 
> terminology is in conformance with the framework). Any other 
> suggestions like this are welcome. 

I noted that there is now some progress in writing down a definition of
`jitter'. 

But we should keep in mind that there are many different quantities of
variation that we will want to measure even in delay, since variation in
latency along a real life path is introduced by many different sources.
Some of these are independent, others are nested heirarchically, etc.

So it will be useful to talk about an overall variation in a data set, as
well as the various components that make up that total. This begs for a
notation.

Beyond the sources of variation, there are different representations of
it, e.g., some sort of range (percentiles), absolute deviations, summed
deviations, squared deviations, etc. 

I suggest that the notation should allow for specifying both the computed
quantity and the source of variation. 

So, for example, if you measure the Mean Squared Error of a sample along
one segment of 19-hop path, you may want to call it something like,
		MSE(hop 2)

But the summed absolute deviations would be:
		SAD(hop 2)

[Note that MSE is standard statistical notation for that quantity, while
SAD is not.]

Andy Scherrer
Statistician
Matrix Information and Directory Services, Inc.
1106 Clayton Lane, Ste. 501W, Austin, TX 78723
http://www.mids.org/
+1-512-451-7602



> Regards, Phil 
> -- 
> -------------------------------------------------------------------------
>        ___o___  Phil Chimento                       tel.: +31 53 4894331
>    ___  | | |   <chimento@cs.utwente.nl>            fax : +31 53 4894524
>   /     | | |   Centre for Telematics and Information Technology  (CTIT)
>   \___  | | |   University of Twente, P.O. Box 217, 7500 AE Enschede, NL
> 
> 
> 



From guest@advanced.org  Tue Aug 10 12:12:59 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA19874
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 12:12:59 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id MAA06397
	for ippm-l@advanced.org; Tue, 10 Aug 1999 12:07:43 -0400 (EDT)
Received: from lri.lri.fr (root@lri.lri.fr [129.175.15.1])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id MAA05391
	for <ippm@advanced.org>; Tue, 10 Aug 1999 12:07:41 -0400 (EDT)
Received: from lri.fr (pc-copri.lri.fr [129.175.13.18])
	by lri.lri.fr (8.9.1a/8.9.1) with ESMTP id SAA19379;
	Tue, 10 Aug 1999 18:07:27 +0200 (MET DST)
Message-ID: <37B04EE8.6AB21D67@lri.fr>
Date: Tue, 10 Aug 1999 18:10:16 +0200
From: =?iso-8859-1?Q?kav=E9?= Salamatian <salamat@lri.fr>
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
X-Accept-Language: en
MIME-Version: 1.0
To: Mike Borella <Mike_Borella@mw.3com.com>
CC: Darryl Veitch <darryl@serc.rmit.edu.au>, ippm@advanced.org
Subject: Re: Loss patterns, Re: minutes for IPPM WG, 13 July 1999
References: <862567C9.0051EFE9.00@mwgate02.mw.3com.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit
Content-Transfer-Encoding: 7bit

Hello,

about this problem of loss patterns, we had also some measure. It seems that for
paths that are shared between large number of transmission, we can use with good
confidence Markovian model such as Elliot-Gilbert  or simple model as noticed by
Darryl, but for links that have not a such an important number of concurrent we can
see very large burst separated by large period of smaller burst. I think that
the burst pattern has things to do with the statistical multiplexing of concurrent
traffics over the link. This was pointed out in simulation by Bolot & al in a paper
"on the  relevance of long range dependence to packet loss in Internet" (I am not
sure of the title of the paper). Then it is not atonishing that the measure of Mike
gives different result from the measure of Darryl. It should be noticed that the
large deployment of RED based router will more bias the Darryl type of observation.

Another interesting things is that we have tried to modelize loss process by some
Semi Markov process, (for the case of large concurrency over the link it reduce to
Markov process). But in most case the Markov chain reduce to only two states. In
fact adding a third, or fourth state does not give much more to the quality of the
model. We cannot says what we see this state aggregation.

Any Idea.

*-------------------------------------------------------------------------------------------------------

Kave Salamatian                                                  salamat@lri.Fr
LRI, URA 410 CNRS
U. Paris SUD
91405 ORSAY
France

Mike Borella wrote:

> How do you mean 'independent'?  Uncorrelated?  Doesn't independent
> mean that the arrival and loss burst processes would both be well modeled as
> geometrically-distributed (at least in the Bernoulli sense)?
>
> And how do the parameters vary over different times of day, days of week,
> time scales and paths?  While the general model of a binary Markov chain
> can be used to model this sort of process, if the parameters are all over the
> map,
> the model may not be very useful in practice.  With any luck the parameter
> values can be ranged based on some physical observations.
>
> I performed a simple analysis of UDP loss patterns a couple of years back on
> 6 months of data across 3 paths.  On all three paths, loss bursts fit a Pareto
> better than anything else I tried.  The interesting thing was that bursts of 10
> or
> less packets lost always had to be modeled differently that bursts of about 15
> or
> greater.  Given the time scales involved (about 30 ms between packets) it seemed
> as if the shorter loss bursts were caused by transient congestion while the
> longer
> bursts (up to and exceeding 1000's of packets) were cause by a more serious
> outage such as downtime or a reboot.
>
> -Mike
>
> Darryl Veitch <darryl@serc.rmit.edu.au> on 08/10/99 02:58:31 AM
>
> Sent by:  Darryl Veitch <darryl@serc.rmit.edu.au>
>
> To:   ippm@advanced.org
> cc:    (Mike Borella/MW/US/3Com)
> Subject:  Loss patterns,     Re: minutes for IPPM WG, 13 July 1999
>
> Hello all,
>
> In a recent mail from Will Leland on the meeting in Oslo, the following comment
> was made:
>
> &&&&&&&&&&&&&&&&&&&&&&&&
> "  There was also a discussion about the fact that a burst loss
> is considered terminated by a single good packet arrival.
> A burst followed by a single good packet and another burst
> would seem to be better characterized as one long burst.
> Will noted that what is being measured could provide the
> raw data for constructing a "less sensitive" metric.
> Would some sort of "errored period" be a meaningful thing
> to define?  "
> &&&&&&&&&&&&&&&&&&&&&&&&
>
> Two students and I did some work last year looking at characterising
> the loss process in detail.
> One of the issues to clarify was precisely this one of, 'ok losses tend to occur
> in small groups, but what about patterns such as  111011 ?
> The results we obtained were quite surprising. We found that the structure of
> losses
> in UDP test packet streams (send out either as Poisson or with constant
> separation)
> was extremely 'independent', in fact looking like an alternating renewal process
> in most cases we observed (~100 experiments over 15 routes in 8 countries).
>
> That is, take two random variables  L (Loss) and A (Arrival).
>  To a good approximation the loss process consisted in simply
> taking sample values alternately from L and A, in an independent fashion.
> This offers the prospect of a simple description of the loss process consisting
> of a discrete probability distribution for L (with most of the mass on 1, 2 or 3
> losses), and an exponential distribution for A.
> With this underlying model the probability of obtaining patterns
> such as 111011, or any others, is easily calculated.
>
> I know it is late days as far as the RFC is concerned, but I think that such a
> description could be of use.   I will be taking up this work in more detail in
> coming months.    Sue Moon presented similar finding at Infocom this year.
>
> The paper appeared in Globecom 98 and can be downloaded from
>  http://www.serc.rmit.edu.au/~darryl     (paper at the bottom)
>
> Darryl Veitch
>
> +-------------------------------+------------------------------------------+
> |      Dr. Darryl Veitch      | Email:   darryl@serc.rmit.edu.au         |
> |      Senior Research Fellow   |                                          |
> |      S.E.R.C.                 | Telephone:                               |
> |      110 Victoria St          |        Direct-   +61 3 9925 4014         |
> |      Carlton 3053             |     Inquiries-   +61 3 9925 4013         |
> |      GPO Box 2476V            |                                          |
> |      Melbourne                | Fax:             +61 3 9925 4094         |
> |      Victoria 3001            |                                          |
> |      Australia                | Web: http://www.serc.rmit.edu.au/~darryl |
> +-------------------------------+------------------------------------------+



From guest@advanced.org  Tue Aug 10 13:34:46 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA21240
	for <ippm-archive@odin.ietf.org>; Tue, 10 Aug 1999 13:34:45 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id NAA05502
	for ippm-l@advanced.org; Tue, 10 Aug 1999 13:28:06 -0400 (EDT)
Received: from tadarida.theasis.com (tadarida.theasis.com [206.9.240.247])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id NAA14036
	for <ippm@advanced.org>; Tue, 10 Aug 1999 13:28:04 -0400 (EDT)
Received: from localhost (andy@localhost)
	by tadarida.theasis.com (8.9.2/8.9.2) with SMTP id MAA18199;
	Tue, 10 Aug 1999 12:27:53 -0500 (CDT)
X-Authentication-Warning: tadarida.theasis.com: andy owned process doing -bs
Date: Tue, 10 Aug 1999 12:27:53 -0500 (CDT)
From: Andy Scherrer <andy@mids.org>
X-Sender: andy@tadarida.theasis.com
To: Mike Borella <Mike_Borella@mw.3com.com>
cc: Darryl Veitch <darryl@serc.rmit.edu.au>, ippm@advanced.org
Subject: Re: Loss patterns, Re: minutes for IPPM WG, 13 July 1999
In-Reply-To: <862567C9.0051EFE9.00@mwgate02.mw.3com.com>
Message-ID: <Pine.GSO.3.96.990810121427.16501T-100000@tadarida.theasis.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


> How do you mean 'independent'?  Uncorrelated?  

Good question; I took it to mean that any packet behaves in a way that is
_not_ more predictable based on information you might have about other
packets in the stream, particularly its neighbors. 

> Doesn't independent
> mean that the arrival and loss burst processes would both be well modeled as
> geometrically-distributed (at least in the Bernoulli sense)?

No, that's sort of backwards. 
As I suggested in an earlier response, many distributions can be used to
model this system, and they all actually assume that samples elements are
independent. 

I believe the property you're thinking of re: Geometric-distributed (and
exponential) data is that of "memorylessness", which is different than
independence. 

Whether that is apt for a model of this sort of system, I'm not yet
convinced. 

> And how do the parameters vary over different times of day, days of week,
> time scales and paths?  

In other words, the parameters of the distribution themselves have
parameters! There has been some work on this, particularly a paper given
at Interface '98 (I forget the author). Modeling changing parameters of,
say, a Poisson distribution does show promise in being useful. 

> While the general model of a binary Markov chain
> can be used to model this sort of process, if the parameters are all over the
> map, the model may not be very useful in practice.  With any luck the
> parameter values can be ranged based on some physical observations.

Please note that elements of a Markov chain are explicitly _not_
independent. In Markov chains, the dependency relationship is merely
restricted, so that, e.g., all the information about the past that is
important to element t is contained in element t-1 (or something that
reduces to that). And so if you know elements t and t-1, you don't gain
anything by considering t-2.

> I performed a simple analysis of UDP loss patterns a couple of years back on
> 6 months of data across 3 paths.  On all three paths, loss bursts fit a Pareto
> better than anything else I tried.  The interesting thing was that bursts of 10
> or
> less packets lost always had to be modeled differently that bursts of about 15
> or
> greater.  Given the time scales involved (about 30 ms between packets) it seemed
> as if the shorter loss bursts were caused by transient congestion while the
> longer
> bursts (up to and exceeding 1000's of packets) were cause by a more serious
> outage such as downtime or a reboot.

Yes, that _is_ very interesting. How much did you learn about what
qualifies for "short" and "long"? I imagine that depends on many different
factors as well.


Andy Scherrer
Statistician
Matrix Information and Directory Services, Inc.
1106 Clayton Lane, Ste. 501W, Austin, TX 78723
http://www.mids.org/
+1-512-451-7602



> -Mike



From matt@advanced.org  Wed Aug 11 03:25:35 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA16702
	for <ippm-archive@odin.ietf.org>; Wed, 11 Aug 1999 03:25:35 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id DAA22958
	for ippm-l@advanced.org; Wed, 11 Aug 1999 03:13:06 -0400 (EDT)
Received: from utrhcs.cs.utwente.nl (utrhcs.cs.utwente.nl [130.89.10.247])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id DAA22962;
	Wed, 11 Aug 1999 03:13:05 -0400 (EDT)
Received: from myrtilos.cs.utwente.nl (myrtilos.cs.utwente.nl [130.89.16.9])
	by utrhcs.cs.utwente.nl (8.9.1/8.9.1) with SMTP id JAA26192;
	Wed, 11 Aug 1999 09:13:00 +0200 (MET DST)
Received: from polya.cs.utwente.nl by myrtilos.cs.utwente.nl (SMI-8.6/csrelay-Sol1.4/RB)
	id JAA02236; Wed, 11 Aug 1999 09:12:57 +0200
Received: from localhost by polya.cs.utwente.nl (8.8.8+Sun/SMI-SVR4)
	id JAA24082; Wed, 11 Aug 1999 09:12:57 +0200 (MET DST)
Message-Id: <199908110712.JAA24082@polya.cs.utwente.nl>
X-Mailer: exmh version 2.0.2 2/24/98
To: Andy Scherrer <andy@mids.org>
Cc: Matthew J Zekauskas <matt@advanced.org>, ippm@advanced.org
Subject: Re: IPDV clock terminology 
In-reply-to: Your message of "Tue, 10 Aug 1999 11:00:51 -0500."
             <Pine.GSO.3.96.990810105107.16501O-100000@tadarida.theasis.com> 
References: <Pine.GSO.3.96.990810105107.16501O-100000@tadarida.theasis.com> 
From: "Philip F. Chimento" <chimento@cs.utwente.nl>
X-Organisation: University of Twente, Dept. of Computer Science
                P.O. Box 217, NL-7500 AE Enschede, The Netherlands
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Wed, 11 Aug 1999 09:12:56 +0200
Sender: chimento@cs.utwente.nl

Hi Andy: 
> 
> I noted that there is now some progress in writing down a definition of
> `jitter'. 
> 
> But we should keep in mind that there are many different quantities of
> variation that we will want to measure even in delay, since variation in
> latency along a real life path is introduced by many different sources.
> Some of these are independent, others are nested heirarchically, etc.
> 
Right; I tried to lay out in Oslo (and its on the slides) what I thought
the sources of the variation are. And yes, you are right, there are a lot 
of different quantities that can be used for "delay variation". 

> So it will be useful to talk about an overall variation in a data set, as
> well as the various components that make up that total. This begs for a
> notation.
> 
> Beyond the sources of variation, there are different representations of
> it, e.g., some sort of range (percentiles), absolute deviations, summed
> deviations, squared deviations, etc. 
> 
I think that these can easily be defined as derivative measures from the 
basic IPDV metric. I think that probably the main criterion for defining them 
will be whether someone thinks that they are useful. In any case, I think 
that a couple of paragraphs are warranted that explain the different 
metrics that could be derived and what their meaning is. 

> I suggest that the notation should allow for specifying both the computed
> quantity and the source of variation. 
> 
> So, for example, if you measure the Mean Squared Error of a sample along
> one segment of 19-hop path, you may want to call it something like,
> 		MSE(hop 2)
>
I don't quite understand the notation here. What does hop 2 signify ? 
 
> 
> Andy Scherrer
> Statistician
> Matrix Information and Directory Services, Inc.
> 1106 Clayton Lane, Ste. 501W, Austin, TX 78723
> http://www.mids.org/
> +1-512-451-7602
> 
Regards, Phil 
-- 
-------------------------------------------------------------------------
       ___o___  Phil Chimento                       tel.: +31 53 4894331
   ___  | | |   <chimento@cs.utwente.nl>            fax : +31 53 4894524
  /     | | |   Centre for Telematics and Information Technology  (CTIT)
  \___  | | |   University of Twente, P.O. Box 217, 7500 AE Enschede, NL




From matt@advanced.org  Wed Aug 11 09:00:42 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA22435
	for <ippm-archive@odin.ietf.org>; Wed, 11 Aug 1999 09:00:42 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id IAA00188
	for ippm-l@advanced.org; Wed, 11 Aug 1999 08:53:26 -0400 (EDT)
Received: from birch.ripe.net (birch.ripe.net [193.0.1.96])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id IAA25133
	for <ippm@advanced.org>; Wed, 11 Aug 1999 08:53:24 -0400 (EDT)
Received: from x49.ripe.net (x49.ripe.net [193.0.1.49])
	by birch.ripe.net (8.8.8/8.8.8) with ESMTP id OAA20146
	for <ippm@advanced.org>; Wed, 11 Aug 1999 14:52:53 +0200 (CEST)
Received: from localhost (henk@localhost)
	by x49.ripe.net (8.8.8/8.8.5) with SMTP id OAA04738
	for <ippm@advanced.org>; Wed, 11 Aug 1999 14:52:53 +0200 (CEST)
X-Authentication-Warning: x49.ripe.net: henk owned process doing -bs
Date: Wed, 11 Aug 1999 14:52:53 +0200 (CEST)
From: "Henk Uijterwaal (RIPE-NCC)" <henk@ripe.net>
To: ippm@advanced.org
Subject: Re: Random comments (was Re: minutes for IPPM WG, 13 July 1999)
In-Reply-To: <Pine.GSO.3.96.990810083524.16501D-100000@tadarida.theasis.com>
Message-ID: <Pine.BSI.3.95.990811141325.4263D-100000@x49.ripe.net>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Tue, 10 Aug 1999, Andy Scherrer wrote:
> I wrote:
> > Matt and Will wrote:

> > > 4. Milestones and futures, Matt Zekauskas and Will Leland (Telcordia).
> > 
> > I don't see it anywhere in the minutes and I'm not even sure anymore if I
> > made the comment in Oslo :-), but I think that we need a (BCP?) document
> > describing how to compare numbers and distributions.  Contrary to to
> > protocols and other binary stuff, this is NOT immediately obvious. 

> > For example:  Suppose two implementations measure a delay over the same
> > path for a time-interval and quote that as (10%, median, 90%).  Now
> > implementation 1 finds (15ms, 30ms, 45ms) and #2 finds (13ms, 26ms, 39ms),
> > does that mean that the values are comparable and that the difference is
> > due to statistical fluctuations and sampling errors _or_ is there a
> > significant difference between the two?

> Adequate comparison needs even more information, including sample sizes.

OK, but then let's at least define which numbers have to be provided (or
have to be constant) in order to be able to compare them.   Besides
comparing implementations, I can also see applications when comparing
today's results for the delays along a certain path against the results
from last week. 

> All statistical comparisons are deemed "significant" or not on the
> basis of variation around the points you want to compare (usually the
> center). While the 10% and 90% values give a range that is useful for
> quantifying variation, the weighting of such a quantity depends on sample
> size. And also on how independent your samples are. 

> And then, defining what qualifies as "significant" is an entirely
> different problem, since at the very least there's a distinction between
> "detectible" with a certain reliability and "important" as far as, say, a
> user or application would be concerned. 

I think we should define "significant" as "detectible", or the point when
there is a certain probability that the different results from two
measurements are caused by a real effect.  A user might want to set a
higher threshold depending on his application.


> > The same goes for distributions.  Look back at the slides I showed in
> > M'polis.  I think we really need some agreed method better than just >
> looking at them, to say that 2 distributions are compatible. 
> 
> To come up with an improvement on just looking at them (noting that humans
> are really good at pattern recognition),

But the human brain quickly tends to get bored.  To compare a handful of
distributions against each other can easily be done by eye.  However, to
compare, say, today's delay patterns for all 40^2 combinations between 40
hosts against yesterday's results, is something that you want to automate. 
Even reducing the 1600 plots for this case to a few % that don't seem to
agree and have to be looked at by a human is already useful. 

> you'd have to identify features
> of the distributions that are important. Once again it's a question of
> weighting: how important are deviations in the tails, vs. deviations in
> the center, or some other region?

Agreed.

Henk

------------------------------------------------------------------------------
Henk Uijterwaal                    Email: henk.uijterwaal@ripe.net
RIPE Network Coordination Centre     WWW: http://www.ripe.net/home/henk
Singel 258                         Phone: +31.20.535-4414,  Fax -4445
1016 AB Amsterdam                   Home: +31.20.4195305
The Netherlands                   Mobile: +31.6.55861746  
------------------------------------------------------------------------------

The Committee (...) was unable to reach a consensus that substantial merit was
lacking. Thus, the appeal was deemed meritorious.          (Orlando NABC #19).



From matt@advanced.org  Wed Aug 11 10:49:54 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA29019
	for <ippm-archive@odin.ietf.org>; Wed, 11 Aug 1999 10:49:53 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id KAA04086
	for ippm-l@advanced.org; Wed, 11 Aug 1999 10:45:59 -0400 (EDT)
Received: from tadarida.theasis.com (tadarida.theasis.com [206.9.240.247])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id KAA26966;
	Wed, 11 Aug 1999 10:45:57 -0400 (EDT)
Received: from localhost (andy@localhost)
	by tadarida.theasis.com (8.9.2/8.9.2) with SMTP id JAA20033;
	Wed, 11 Aug 1999 09:45:46 -0500 (CDT)
X-Authentication-Warning: tadarida.theasis.com: andy owned process doing -bs
Date: Wed, 11 Aug 1999 09:45:46 -0500 (CDT)
From: Andy Scherrer <andy@mids.org>
X-Sender: andy@tadarida.theasis.com
To: "Philip F. Chimento" <chimento@cs.utwente.nl>
cc: Matthew J Zekauskas <matt@advanced.org>, ippm@advanced.org
Subject: Re: IPDV clock terminology 
In-Reply-To: <199908110712.JAA24082@polya.cs.utwente.nl>
Message-ID: <Pine.GSO.3.96.990811093725.16501a-100000@tadarida.theasis.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


> > Beyond the sources of variation, there are different representations of
> > it, e.g., some sort of range (percentiles), absolute deviations, summed
> > deviations, squared deviations, etc. 
> > 
> I think that these can easily be defined as derivative measures from the 
> basic IPDV metric. I think that probably the main criterion for defining them 
> will be whether someone thinks that they are useful. In any case, I think 
> that a couple of paragraphs are warranted that explain the different 
> metrics that could be derived and what their meaning is. 

OK -- mainly what I was getting at was it might be useful to think on how
to allow for the specification of many sources of variation (those deemed
important) in a consistent way, but not really to try and list & define
every possible important source.

> > I suggest that the notation should allow for specifying both the computed
> > quantity and the source of variation. 
> > 
> > So, for example, if you measure the Mean Squared Error of a sample along
> > one segment of 19-hop path, you may want to call it something like,
> > 		MSE(hop 2)
> >
> I don't quite understand the notation here. What does hop 2 signify ? 

I omitted a sentence of explanation... 
(hop2) merely specifies the entity to which the quantity applies, or the
source of variation. I was thinking of "hop 2" as the second segment on a
path -- just a vague and arbitrary example. It probably would have been
better for me to use something like,
		MSE(router 1.2.3.4)  
indicating that part of the total variation which was due to the influence
of that router. 

Hopefully that's a little clearer.

Andy


> Regards, Phil 
> -- 
> -------------------------------------------------------------------------
>        ___o___  Phil Chimento                       tel.: +31 53 4894331
>    ___  | | |   <chimento@cs.utwente.nl>            fax : +31 53 4894524
>   /     | | |   Centre for Telematics and Information Technology  (CTIT)
>   \___  | | |   University of Twente, P.O. Box 217, 7500 AE Enschede, NL
> 
> 
> 



From matt@advanced.org  Thu Aug 12 06:42:14 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA28215
	for <ippm-archive@odin.ietf.org>; Thu, 12 Aug 1999 06:42:14 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id GAA10751
	for ippm-l@advanced.org; Thu, 12 Aug 1999 06:29:30 -0400 (EDT)
Received: from universe.serc.rmit.edu.au (universe.serc.rmit.edu.au [131.170.42.12])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id GAA17839
	for <ippm@advanced.org>; Thu, 12 Aug 1999 06:29:25 -0400 (EDT)
Received: (from darryl@localhost)
	by universe.serc.rmit.edu.au (8.8.5/8.8.5) id UAA26096;
	Thu, 12 Aug 1999 20:28:25 +1000 (EST)
Date: Thu, 12 Aug 1999 20:28:25 +1000 (EST)
From: Darryl Veitch <darryl@serc.rmit.edu.au>
Message-Id: <199908121028.UAA26096@universe.serc.rmit.edu.au>
To: andy@mids.org, mike_borella@mw.3com.com
Subject: Re: Loss patterns,     
Cc: ippm@advanced.org
X-Sun-Charset: US-ASCII


Let me try to answer these questions briefly myself.

From: "Mike Borella" <Mike_Borella@mw.3com.com>

**  How do you mean 'independent'?  Uncorrelated? 
I am suggesting a model where they are independent, based on an analysis which looks
at covariances.   Of course they are not truly independent, but the idea is that if
such a simple model can be of use, let's be glad..


**  Doesn't independent
**  mean that the arrival and loss burst processes would both be well modeled as
**  geometrically-distributed (at least in the Bernoulli sense)?
If I understand what you mean, yes.  In this model if you isolate either
of the losses or the arrival 'bursts' as discrete time series  L(k) and A(k) 
where L(k) is the # of losses in the kth loss burst, and similarly for A, then
both of these would be iid processes.


   
**  And how do the parameters vary over different times of day, days of week,
**  time scales and paths?  While the general model of a binary Markov chain
**  can be used to model this sort of process, if the parameters are all over the
**  map,
**  the model may not be very useful in practice.  With any luck the parameter
**  values can be ranged based on some physical observations.

More work needs to be done here.  What we did was to first attempt to isolate
long segments of traces which appeared stationary.  Although parameters varied,
for such segments the model structure seemed to apply well.  
The values of the parameters 
changes though.  Naturally a parameterl-ess model, 'universal traffic', would
be a lot to ask for.. 
 As for 'parameters of parameters', this is what we do all the time really.

  
**  I performed a simple analysis of UDP loss patterns a couple of years back on
**  6 months of data across 3 paths.  On all three paths, loss bursts fit a Pareto
**  better than anything else I tried.  The interesting thing was that bursts of 10
**  or
**  less packets lost always had to be modeled differently that bursts of about 15
**  or
**  greater.  Given the time scales involved (about 30 ms between packets) it seemed
**  as if the shorter loss bursts were caused by transient congestion while the
**  longer
**  bursts (up to and exceeding 1000's of packets) were cause by a more serious
**  outage such as downtime or a reboot.


That sounds reasonable.  Our 'stationary' hunting would tend to exclude these 
worst excesses, which could be modelled separately, or but adding a pareto tail
of low mass to the loss burst distribution, or perhaps just some mass at a single
large value-  depends what you need to model.
  


From: Andy Scherrer <andy@mids.org>

**  > That is, take two random variables L (Loss) and A (Arrival). 
**  >  To a good approximation the loss process consisted in simply taking
**  > sample values alternately from L and A, in an independent fashion. 
**  
**  But in what proportions?  Of course, that's a question you ask of the
**  data, but in a sequence whose beginning and end are difficult to define,
**  the time scale becomes an issue. And so while loss of one packet may not
**  be dependent on loss of any other packet, there is dependency on time
**  scale, since it will probably matter if you're looking at such proportions
**  in 3 sec bursts vs. 30 sec or 3 min bursts.

I am not sure what you mean by 'proportions'.  When I say alternate I mean that.
In the analysis all proportions were looked at via the cross correlation of the
two burst series, which was very low.

 
**  > This offers the prospect of a simple description of the loss process
**  > consisting of a discrete probability distribution for L (with most of
**  > the mass on 1, 2 or 3 losses), and an exponential distribution for A. 
**  
**  But note that using a simple univariate exponential distribution for A
**  implies a Poisson distribution. This group has discussed problems with
**  Poisson modelling of these data. That is not to say it shouldn't perhaps
**  be discussed more, perhaps with some thought to heirarchical modeling of
**  the Poisson parameter (lambda) -- that would be one approach for
**  addressing the time issue I mentioned above. 

The overall process I describe is not Poisson, although it is short range 
dependent of course.  (paradoxically, we did see some evidence of LRD, I believe
that it is there sometimes but is of low 'power'  (low mass in tail).


 
**  > With this underlying model the probability of obtaining patterns such as
**  > 111011, or any others, is easily calculated. 
**  
**  Well, I submit that there are actually a suite of underlying models to
**  consider. A binary outcome such as your example can result from one of
**  many distributions, each of which thinks of the "question" in a different
**  way. 
**  
**  For example, if you think of it as Binomial problem (a series of
**  Bernoulli[0-1] trials), you have a fixed, known number of trials (N), 
**  and model the proportion out of N that are 1 (or 0).
**  
**  However, another way of looking at the _same_ data could be "How many
**  trials (packets) do I see before I observe a 1"? Doing this repeatedly
**  might give a set of data consisting of sequence lengths.
**  
**   ...which is different from "How many trials (packets) do I observe before
**  I can count some number C of 1s?" This can also give data consisting of
**  sequence lengths, but counted differently. 

I believe that the framework I described, ie packet k is missing or not, is in some
sense the 'primitive process' from which all these alternately viewpoints can be
deduced.   The converse however is not always true.

 

**  Hello,
**  
**  about this problem of loss patterns, we had also some measure. It seems that for
**  paths that are shared between large number of transmission, we can use with good
**  confidence Markovian model such as Elliot-Gilbert  or simple model as noticed by
**  Darryl, but for links that have not a such an important number of concurrent we can
**  see very large burst separated by large period of smaller burst. I think that
**  the burst pattern has things to do with the statistical multiplexing of concurrent
**  traffics over the link. This was pointed out in simulation by Bolot & al in a paper
**  "on the  relevance of long range dependence to packet loss in Internet" (I am not
**  sure of the title of the paper). Then it is not atonishing that the measure of Mike
**  gives different result from the measure of Darryl. It should be noticed that the
**  large deployment of RED based router will more bias the Darryl type of observation.
**  
**  Another interesting things is that we have tried to modelize loss process by some
**  Semi Markov process, (for the case of large concurrency over the link it reduce to
**  Markov process). But in most case the Markov chain reduce to only two states. In
**  fact adding a third, or fourth state does not give much more to the quality of the
**  model. We cannot says what we see this state aggregation.
**  
**  Any Idea.
**  
**  *-------------------------------------------------------------------------------------------------------
**  
**  Kave Salamatian                                                  salamat@lri.Fr
**  LRI, URA 410 CNRS
**  U. Paris SUD
**  91405 ORSAY
**  France
 

I agree that on links with low multiplexing, that extreme events are more likely
and other models would be better.  On the other hand large bursts could be modelled,
perhaps successfully, by adding a new parameter which gives a reasonable mass to 
some large burst size.  Of course this flies in the face of the ethos of scaling 
(fractal) behaviour, god forbid that I should be saying this myself!


Darryl 


+-------------------------------+------------------------------------------+
|      Dr. Darryl Veitch 	| Email:   darryl@serc.rmit.edu.au         |
|      Senior Research Fellow   |                                          |
|      S.E.R.C.                 | Telephone:                               |
|      110 Victoria St          |        Direct-   +61 3 9925 4014         |
|      Carlton 3053             |     Inquiries-   +61 3 9925 4013         |
|      GPO Box 2476V            |                                          |
|      Melbourne                | Fax:             +61 3 9925 4094         |
|      Victoria 3001            |                                          |
|      Australia                | Web: http://www.serc.rmit.edu.au/~darryl |
+-------------------------------+------------------------------------------+




From matt@advanced.org  Fri Aug 13 03:47:12 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id DAA13769
	for <ippm-archive@odin.ietf.org>; Fri, 13 Aug 1999 03:47:12 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id DAA07930
	for ippm-l@advanced.org; Fri, 13 Aug 1999 03:38:59 -0400 (EDT)
Received: from daffy.ee.lbl.gov (daffy.ee.lbl.gov [131.243.1.31])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id DAA25117
	for <ippm@advanced.org>; Fri, 13 Aug 1999 03:38:57 -0400 (EDT)
Received: (from vern@localhost)
	by daffy.ee.lbl.gov (8.9.2/8.9.2) id AAA28708;
	Fri, 13 Aug 1999 00:38:50 -0700 (PDT)
Message-Id: <199908130738.AAA28708@daffy.ee.lbl.gov>
To: Darryl Veitch <darryl@serc.rmit.edu.au>
Cc: andy@mids.org, mike_borella@mw.3com.com, ippm@advanced.org
Subject: Re: Loss patterns, 
In-reply-to: Your message of Thu, 12 Aug 1999 20:28:25 +1000.
Date: Fri, 13 Aug 1999 00:38:50 PDT
From: Vern Paxson <vern@ee.lbl.gov>

Just a note of something to keep in mind: while studies have generally
found loss patterns independent for packets spaced hundreds of msec or
more apart, on finer time scales losses can be strongly correlated.  E.g.,
in my packet dynamics study the probability that a TCP packet is lost given
that its predecessor was lost was 5-8 times larger than if its predecessor
was not lost.  This will likely change with wide deployment of RED: but
it argues for caution when applying models based on independence of loss.

		Vern



From matt@advanced.org  Fri Aug 13 05:00:05 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA14557
	for <ippm-archive@odin.ietf.org>; Fri, 13 Aug 1999 05:00:04 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id EAA22646
	for ippm-l@advanced.org; Fri, 13 Aug 1999 04:54:34 -0400 (EDT)
Received: from universe.serc.rmit.edu.au (universe.serc.rmit.edu.au [131.170.42.12])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id EAA24332
	for <ippm@advanced.org>; Fri, 13 Aug 1999 04:54:30 -0400 (EDT)
Received: (from darryl@localhost)
	by universe.serc.rmit.edu.au (8.8.5/8.8.5) id SAA28460;
	Fri, 13 Aug 1999 18:54:24 +1000 (EST)
Date: Fri, 13 Aug 1999 18:54:24 +1000 (EST)
From: Darryl Veitch <darryl@serc.rmit.edu.au>
Message-Id: <199908130854.SAA28460@universe.serc.rmit.edu.au>
To: vern@ee.lbl.gov
Subject: Re: Loss patterns,
Cc: ippm@advanced.org
X-Sun-Charset: US-ASCII

Whilst not wishing to imply that the model I suggested, based on UDP observations, is
relevant for TCP,  note that it is Not an iid model, and is rich enough to 
include the conditional probabilities you mention.

Essentially the probability of loss given that one is within an arrival burst is 
governed by the residual life distribution relating to the distribution A, whereas 
if one assumes that the previous packet was lost, then one is within a loss burst by
definition and the residual life of L governs the probability.  They are therefore not
the same in general.    Although the model has a lot of 'independence' in it, it is
far from an iid model, being an alternating renewal process. 


Darryl




**  From matt@advanced.org Fri Aug 13 18:32 EST 1999
**  To: Darryl Veitch <darryl@serc.rmit.edu.au>
**  Cc: andy@mids.org, mike_borella@mw.3com.com, ippm@advanced.org
**  Subject: Re: Loss patterns, 
**  Date: Fri, 13 Aug 1999 00:38:50 PDT
**  From: Vern Paxson <vern@ee.lbl.gov>
**  
**  Just a note of something to keep in mind: while studies have generally
**  found loss patterns independent for packets spaced hundreds of msec or
**  more apart, on finer time scales losses can be strongly correlated.  E.g.,
**  in my packet dynamics study the probability that a TCP packet is lost given
**  that its predecessor was lost was 5-8 times larger than if its predecessor
**  was not lost.  This will likely change with wide deployment of RED: but
**  it argues for caution when applying models based on independence of loss.
**  
**  		Vern
**  
 



From matt@advanced.org  Sat Aug 14 02:25:18 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id CAA00900
	for <ippm-archive@odin.ietf.org>; Sat, 14 Aug 1999 02:25:17 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id CAA22439
	for ippm-l@advanced.org; Sat, 14 Aug 1999 02:20:34 -0400 (EDT)
Received: from daffy.ee.lbl.gov (daffy.ee.lbl.gov [131.243.1.31])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id CAA24727
	for <ippm@advanced.org>; Sat, 14 Aug 1999 02:20:32 -0400 (EDT)
Received: (from vern@localhost)
	by daffy.ee.lbl.gov (8.9.2/8.9.2) id XAA02805;
	Fri, 13 Aug 1999 23:20:27 -0700 (PDT)
Message-Id: <199908140620.XAA02805@daffy.ee.lbl.gov>
To: Darryl Veitch <darryl@serc.rmit.edu.au>
Cc: ippm@advanced.org
Subject: Re: Loss patterns,
Date: Fri, 13 Aug 1999 23:20:26 PDT
From: Vern Paxson <vern@ee.lbl.gov>

> Whilst not wishing to imply that the model I suggested, based on UDP
> observations, is relevant for TCP,  note that it is Not an iid model, and
> is rich enough to include the conditional probabilities you mention ...
> ... being an alternating renewal process.

Ah!, got it - that's a model I like, it strikes me as a good approximation.
Note that in my packet dynamics study, the distribution for ON-times for
ack loss periods had a clear Pareto tail with infinite variance - definitely
an important property to keep.

		Vern



From matt@advanced.org  Sun Aug 29 17:52:43 1999
Received: from betelgeuse.advanced.org (betelgeuse.advanced.org [199.222.103.10])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA02604
	for <ippm-archive@odin.ietf.org>; Sun, 29 Aug 1999 17:52:43 -0400 (EDT)
Received: (from guest@localhost)
	by betelgeuse.advanced.org (8.9.3/8.9.1) id RAA20621
	for ippm-l@advanced.org; Sun, 29 Aug 1999 17:42:56 -0400 (EDT)
Received: from mailhost.auckland.ac.nz (mailhost.auckland.ac.nz [130.216.1.4])
	by betelgeuse.advanced.org (8.9.3/8.9.1) with ESMTP id RAA21025
	for <ippm@advanced.org>; Sun, 29 Aug 1999 17:42:52 -0400 (EDT)
Received: from n.browlee5.itss.auckland.ac.nz (n.brownlee5.itss.auckland.ac.nz [130.216.4.79])
	by mailhost.auckland.ac.nz (8.9.2/8.9.2/8.9.2-ua) with SMTP id JAA22840
	for <ippm@advanced.org>; Mon, 30 Aug 1999 09:42:50 +1200 (NZST)
From: Nevil Brownlee <n.brownlee@auckland.ac.nz>
Sender: jbro111@auckland.ac.nz
To: ippm@advanced.org
Subject: Conference: Call for Papers
Message-ID: <SIMEON.9908300904.B@n.auckland.ac.nz>
Date: Mon, 30 Aug 1999 09:43:04 +1200 (New Zealand Standard Time)
Priority: NORMAL
X-Mailer: Simeon for Win32 Version 4.1.4 Build (40)
X-Authentication: none
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII



             Passive & Active Measurement: PAM-2000

                         Call for papers


A workshop on passive and active measurement techniques for high speed
computer networks and the Internet


           University of Waikato, Hamilton, New Zealand
                       April 3 and 4, 2000

                Abstracts due: 15 November 1999
              Full papers due: 15 February 2000


Conference Chair: Professor Ian Graham, University of Waikato

Program Committee:
Nevil Brownlee (chair)   The University of Auckland, New Zealand
Scott Bradner            Harvard University, U.S.A.
Hans-Werner Braun        NLANR, U.S.A.
Randy Bush               Verio, U.S.A.        
John Cleary              University of Waikato, New Zealand
Geoff Huston             Telstra, Australia
Kitamura Yasuichi        CRL/APAN, Japan
Tracie Monk              CAIDA, U.S.A.
Alastair Reynolds        Hewlett Packard, Scotland
Matthew Robinson         RIPE Test Traffic WG, U.K.
Henk Uijterwaal          RIPE NCC, Netherlands
Matt Zekauskas           Advanced Network & Services, U.S.A.    


As the Internet has grown over the last decade the need for precise
measurement of network traffic has become steadily more apparent; most
of today's Internet Service Providers and many of their large network
customers are collecting and analysing traffic data for the purposes of
performance monitoring, network engineering and cost recovery, but the
engineering quality of these measurements vary. 

A steadily growing number of research groups have been
working in the areas of
 - Active Measurements, i.e. sending test packets and observing
      their progress through the Internet
 - Passive Measurements, i.e. observing actual traffic on 'live' 
      networks
 - Performance Metrics, i.e. developing measures or indicators
      which can be used to characterise traffic behaviour
 - Traffic Statistics, i.e. attempting to understand and develop
      models of 'real' Internet traffic
 - Visualisation, i.e. finding effective ways to display
      what's happening in a network

One such group is the Waikato Applied Network Dynamics (WAND) group,
based at Waikato University in New Zealand.  The WAND group will be
local host to a workshop to be held in New Zealand early in April 2000,
following the IETF meeting in Adelaide. 


Papers are invited from the research, provider and any other communities
on topics in the areas above, or any other area of network traffic
measurement.  Papers reporting practical experience of measurement
and/or analysis are especially welcome, in particular:

 1) 'Experience' papers, which talk about practical uses of
    measurement, especially in large networks are strongly encouraged.
 2) Papers on modelling would also be very welcome, especially if
    they're backed by measurements of real traffic.
 3) More 'academic' papers are fine too, but please emphasise
    the 'practical' side of the measurements.


Extended abstracts (about 500 words) must be submitted by e-mail to
pam2000@cs.waikato.ac.nz by 15 November 1999. Abstracts will be reviewed
and acceptance notified by 15 December 1999 or earlier. 

Papers, maximum 15 printed pages, must be submitted electronically by 15
February 2000. Details of acceptable paper formats will be published on
the conference web site, pam2000.cs.waikato.ac.nz. 


Equipment and software demonstrations are also welcomed. Proposals for
demonstrations should be sent by email to the conference chair at
pam2000@cs.waikato.ac.nz by 15 December 1999.


Hamilton is a city of some 100,000 inhabitants, situated 100 kilometres
(80 miles) south of New Zealand's largest city, Auckland. The weather in
April (our fall in New Zealand) is usually still warm and not too wet.

This workshop follows immediately after the 47th IETF meeting in
Adelaide, Australia (March 27-31, 2000).  Travel from Adelaide through
Auckland to the US is possible with United, Air New Zealand and Quantas
(and possibly other airlines).  Local travel and tourism can be arranged
through the workshop's travel agents in Hamilton; Calder and Lawson,
http://www.cnl.co.nz. 


The workshop will be held at the Novotel Tainui in Hamilton.
Special room rates are available for the conference,
and the conference fee will be about $US 150 ($NZ 250).
A paper copy of the workshop proceedings is included in the
conference fee.

For further information see http://www2.wave.co.nz/~novotel/hnindex.html

The workshop's web page is http://pam2000.cs.waikato.ac.nz

________________________________________________________________________