From timothy.plunkett at navy.mil  Wed Apr  4 14:01:59 2007
From: timothy.plunkett at navy.mil (Plunkett, Timothy R CIV NSWCDD, W13)
Date: Wed, 4 Apr 2007 16:01:59 -0500
Subject: [ntpwg] questions about current NTP protocol draft
Message-ID: <88ED8F2A922CE84986619E21B45CD4DA02349D94@NAEAMILLEX06VA.nadsusea.nads.navy.mil>

I am currently editing the XML version of the latest NTPv4 protocol
draft.  I noticed the following two things that I was hoping that
someone could clarify for me.

1.  Should the equation for capepsilon in section 10 be for a variable
called EPSILON instead?  I know that there has been an effort to remove
all the "cap" variables and I think this is a leftover.

2.  Should the equation for PSI in section 11.2.3 have a square root
function in it?  Seems like PSI should equal SQRT [(PSI_s)^2 +
(PSI_p)^2].  If preferred I could also raise it to the one half power as
well.

Thanks,

Tim

***********************************************************
Timothy R. Plunkett
NSWC-DD, Code W13
Building 1500 Room A122A
17214 AVENUE B SUITE 100
DAHLGREN, VA 22448-5147

(540) 653 - 1090 (phone at Dahlgren)
(540) 653 - 8673 (fax at Dahlgren)


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070404/d09985c1/attachment.html 

From GDowd at symmetricom.com  Wed Apr  4 14:22:10 2007
From: GDowd at symmetricom.com (Greg Dowd)
Date: Wed, 4 Apr 2007 14:22:10 -0700
Subject: [ntpwg] questions about current NTP protocol draft
In-Reply-To: <88ED8F2A922CE84986619E21B45CD4DA02349D94@NAEAMILLEX06VA.nadsusea.nads.navy.mil>
Message-ID: <4B73D9EBE2327E40B4D69822D7C45802068D08AF@sjmail1.symmetricom.com>

I'm honestly not EXACTLY sure what version is what in the xml domain but
I'll vote.
 
1) I don't see capepsilon in the non xml version.  Can you quote the
string or equation here?  In any case, capepsilon is root dispersion.
If the notes or context allows you to determine that the variable in
question is indeed root dispersion then EPSILON would be correct in the
latest version.
 
2) Yes, sqrt of the sum of the squares.  
 
 

Greg Dowd 
gdowd at symmetricom dot com (antispam format) 
Symmetricom, Inc. 
www.symmetricom.com 
"The current implementation is non-obvious and may need to be improved."




 


________________________________

From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org
[mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org] On Behalf
Of Plunkett, Timothy R CIV NSWCDD, W13
Sent: Wednesday, April 04, 2007 2:02 PM
To: ntpwg at lists.ntp.isc.org
Subject: [ntpwg] questions about current NTP protocol draft



I am currently editing the XML version of the latest NTPv4 protocol
draft.  I noticed the following two things that I was hoping that
someone could clarify for me.

1.  Should the equation for capepsilon in section 10 be for a variable
called EPSILON instead?  I know that there has been an effort to remove
all the "cap" variables and I think this is a leftover.

2.  Should the equation for PSI in section 11.2.3 have a square root
function in it?  Seems like PSI should equal SQRT [(PSI_s)^2 +
(PSI_p)^2].  If preferred I could also raise it to the one half power as
well.

Thanks, 

Tim 

*********************************************************** 
Timothy R. Plunkett 
NSWC-DD, Code W13 
Building 1500 Room A122A 
17214 AVENUE B SUITE 100 
DAHLGREN, VA 22448-5147 

(540) 653 - 1090 (phone at Dahlgren) 
(540) 653 - 8673 (fax at Dahlgren) 


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070404/23811da8/attachment.html 

From karen.odonoghue at navy.mil  Wed Apr  4 14:55:35 2007
From: karen.odonoghue at navy.mil (Odonoghue, Karen F CIV NSWCDD, W13)
Date: Wed, 4 Apr 2007 16:55:35 -0500
Subject: [ntpwg] questions about current NTP protocol draft
References: <4B73D9EBE2327E40B4D69822D7C45802068D08AF@sjmail1.symmetricom.com>
Message-ID: <2E44C2268B59514C89AE2CB514D15ACAD2E272@NAEAMILLEX01VA.nadsusea.nads.navy.mil>

I asked Tim to check the reformatted xml to make sure nothing got corrupted in the process. 
The equation in question is: 
 
Karen
 
   In the next step the shift register stages are copied to a temporary
   list and the list sorted by increasing delta.  Let i index the stages
   starting with the lowest delta.  If the first tuple epoch t_0 is not
   later than the last valid sample epoch p.t, the routine exits without
   affecting the current peer variables.  Otherwise, let epsilon_i be
   the dispersion of the ith entry, then
                     i=n-1
                     ---     epsilon_i
      capepsilon =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0

   is the peer dispersion p.disp.  Note the overload of epsilon, whether
   input to the clock filter or output, the meaning should be clear from
   context.


________________________________

From: ntpwg-bounces+karen.odonoghue=navy.mil at lists.ntp.isc.org on behalf of Greg Dowd
Sent: Wed 4/4/2007 4:22 PM
To: Plunkett, Timothy R CIV NSWCDD, W13; ntpwg at lists.ntp.isc.org
Subject: Re: [ntpwg] questions about current NTP protocol draft


I'm honestly not EXACTLY sure what version is what in the xml domain but I'll vote.
 
1) I don't see capepsilon in the non xml version.  Can you quote the string or equation here?  In any case, capepsilon is root dispersion.  If the notes or context allows you to determine that the variable in question is indeed root dispersion then EPSILON would be correct in the latest version.
 
2) Yes, sqrt of the sum of the squares.  
 
 

Greg Dowd 
gdowd at symmetricom dot com (antispam format) 
Symmetricom, Inc. 
www.symmetricom.com 
"The current implementation is non-obvious and may need to be improved." 



 


________________________________

From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org [mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org] On Behalf Of Plunkett, Timothy R CIV NSWCDD, W13
Sent: Wednesday, April 04, 2007 2:02 PM
To: ntpwg at lists.ntp.isc.org
Subject: [ntpwg] questions about current NTP protocol draft



I am currently editing the XML version of the latest NTPv4 protocol draft.  I noticed the following two things that I was hoping that someone could clarify for me.

1.  Should the equation for capepsilon in section 10 be for a variable called EPSILON instead?  I know that there has been an effort to remove all the "cap" variables and I think this is a leftover.

2.  Should the equation for PSI in section 11.2.3 have a square root function in it?  Seems like PSI should equal SQRT [(PSI_s)^2 + (PSI_p)^2].  If preferred I could also raise it to the one half power as well.

Thanks, 

Tim 

*********************************************************** 
Timothy R. Plunkett 
NSWC-DD, Code W13 
Building 1500 Room A122A 
17214 AVENUE B SUITE 100 
DAHLGREN, VA 22448-5147 

(540) 653 - 1090 (phone at Dahlgren) 
(540) 653 - 8673 (fax at Dahlgren) 


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070404/81a8d033/attachment.html 

From GDowd at symmetricom.com  Wed Apr  4 15:09:31 2007
From: GDowd at symmetricom.com (Greg Dowd)
Date: Wed, 4 Apr 2007 15:09:31 -0700
Subject: [ntpwg] questions about current NTP protocol draft
In-Reply-To: <2E44C2268B59514C89AE2CB514D15ACAD2E272@NAEAMILLEX01VA.nadsusea.nads.navy.mil>
Message-ID: <4B73D9EBE2327E40B4D69822D7C45802068D08B0@sjmail1.symmetricom.com>

That's peer dispersion epsilon, as opposed to root dispersion
capepsilon.    The text in the graphic shows the correct epsilon.
 

Greg Dowd 
gdowd at symmetricom dot com (antispam format) 
Symmetricom, Inc. 
www.symmetricom.com 
"The current implementation is non-obvious and may need to be improved."




 

________________________________

From: Odonoghue, Karen F CIV NSWCDD, W13
[mailto:karen.odonoghue at navy.mil] 
Sent: Wednesday, April 04, 2007 2:56 PM
To: Greg Dowd; Plunkett, Timothy R CIV NSWCDD, W13;
ntpwg at lists.ntp.isc.org
Subject: RE: [ntpwg] questions about current NTP protocol draft


I asked Tim to check the reformatted xml to make sure nothing got
corrupted in the process. 
The equation in question is: 
 
Karen
 
   In the next step the shift register stages are copied to a temporary
   list and the list sorted by increasing delta.  Let i index the stages
   starting with the lowest delta.  If the first tuple epoch t_0 is not
   later than the last valid sample epoch p.t, the routine exits without
   affecting the current peer variables.  Otherwise, let epsilon_i be
   the dispersion of the ith entry, then
                     i=n-1
                     ---     epsilon_i
      capepsilon =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0

   is the peer dispersion p.disp.  Note the overload of epsilon, whether
   input to the clock filter or output, the meaning should be clear from
   context.


________________________________

From: ntpwg-bounces+karen.odonoghue=navy.mil at lists.ntp.isc.org on behalf
of Greg Dowd
Sent: Wed 4/4/2007 4:22 PM
To: Plunkett, Timothy R CIV NSWCDD, W13; ntpwg at lists.ntp.isc.org
Subject: Re: [ntpwg] questions about current NTP protocol draft


I'm honestly not EXACTLY sure what version is what in the xml domain but
I'll vote.
 
1) I don't see capepsilon in the non xml version.  Can you quote the
string or equation here?  In any case, capepsilon is root dispersion.
If the notes or context allows you to determine that the variable in
question is indeed root dispersion then EPSILON would be correct in the
latest version.
 
2) Yes, sqrt of the sum of the squares.  
 
 

Greg Dowd 
gdowd at symmetricom dot com (antispam format) 
Symmetricom, Inc. 
www.symmetricom.com 
"The current implementation is non-obvious and may need to be improved."




 


________________________________

From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org
[mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org] On Behalf
Of Plunkett, Timothy R CIV NSWCDD, W13
Sent: Wednesday, April 04, 2007 2:02 PM
To: ntpwg at lists.ntp.isc.org
Subject: [ntpwg] questions about current NTP protocol draft



I am currently editing the XML version of the latest NTPv4 protocol
draft.  I noticed the following two things that I was hoping that
someone could clarify for me.

1.  Should the equation for capepsilon in section 10 be for a variable
called EPSILON instead?  I know that there has been an effort to remove
all the "cap" variables and I think this is a leftover.

2.  Should the equation for PSI in section 11.2.3 have a square root
function in it?  Seems like PSI should equal SQRT [(PSI_s)^2 +
(PSI_p)^2].  If preferred I could also raise it to the one half power as
well.

Thanks, 

Tim 

*********************************************************** 
Timothy R. Plunkett 
NSWC-DD, Code W13 
Building 1500 Room A122A 
17214 AVENUE B SUITE 100 
DAHLGREN, VA 22448-5147 

(540) 653 - 1090 (phone at Dahlgren) 
(540) 653 - 8673 (fax at Dahlgren) 


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070404/ec825b6c/attachment-0001.html 

From timothy.plunkett at navy.mil  Wed Apr  4 15:16:51 2007
From: timothy.plunkett at navy.mil (Plunkett, Timothy R CIV NSWCDD, W13)
Date: Wed, 4 Apr 2007 17:16:51 -0500
Subject: [ntpwg] questions about current NTP protocol draft
In-Reply-To: <4B73D9EBE2327E40B4D69822D7C45802068D08B0@sjmail1.symmetricom.com>
References: <2E44C2268B59514C89AE2CB514D15ACAD2E272@NAEAMILLEX01VA.nadsusea.nads.navy.mil>
	<4B73D9EBE2327E40B4D69822D7C45802068D08B0@sjmail1.symmetricom.com>
Message-ID: <88ED8F2A922CE84986619E21B45CD4DA02349D96@NAEAMILLEX06VA.nadsusea.nads.navy.mil>

Greg,

But what I am saying is that 


                     i=n-1
                     ---     epsilon_i
      capepsilon =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0

should be changed to be instead


                     i=n-1
                     ---     epsilon_i
         EPSILON =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0
or maybe the capepsilon should only be epsilon, but I am sure that
"capepsilon" should no longer be there.  There are no other variables in
the document that start with "cap".  They have been replaced with an ALL
CAPS version of them.

Tim


***********************************************************
Timothy R. Plunkett
NSWC-DD, Code W13
Building 1500 Room A122A
17214 AVENUE B SUITE 100
DAHLGREN, VA 22448-5147

(540) 653 - 1090 (phone at Dahlgren)
(540) 653 - 8673 (fax at Dahlgren)


-----Original Message-----
From: Greg Dowd [mailto:GDowd at symmetricom.com] 
Sent: Wednesday, April 04, 2007 18:10
To: Odonoghue, Karen F CIV NSWCDD, W13; Plunkett, Timothy R CIV NSWCDD,
W13; ntpwg at lists.ntp.isc.org
Subject: RE: [ntpwg] questions about current NTP protocol draft

That's peer dispersion epsilon, as opposed to root dispersion
capepsilon.    The text in the graphic shows the correct epsilon.
 

Greg Dowd 
gdowd at symmetricom dot com (antispam format) 
Symmetricom, Inc. 
www.symmetricom.com 
"The current implementation is non-obvious and may need to be improved."




 

________________________________

From: Odonoghue, Karen F CIV NSWCDD, W13
[mailto:karen.odonoghue at navy.mil] 
Sent: Wednesday, April 04, 2007 2:56 PM
To: Greg Dowd; Plunkett, Timothy R CIV NSWCDD, W13;
ntpwg at lists.ntp.isc.org
Subject: RE: [ntpwg] questions about current NTP protocol draft


I asked Tim to check the reformatted xml to make sure nothing got
corrupted in the process. 
The equation in question is: 
 
Karen
 
   In the next step the shift register stages are copied to a temporary
   list and the list sorted by increasing delta.  Let i index the stages
   starting with the lowest delta.  If the first tuple epoch t_0 is not
   later than the last valid sample epoch p.t, the routine exits without
   affecting the current peer variables.  Otherwise, let epsilon_i be
   the dispersion of the ith entry, then
                     i=n-1
                     ---     epsilon_i
      capepsilon =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0

   is the peer dispersion p.disp.  Note the overload of epsilon, whether
   input to the clock filter or output, the meaning should be clear from
   context.


________________________________

From: ntpwg-bounces+karen.odonoghue=navy.mil at lists.ntp.isc.org on behalf
of Greg Dowd
Sent: Wed 4/4/2007 4:22 PM
To: Plunkett, Timothy R CIV NSWCDD, W13; ntpwg at lists.ntp.isc.org
Subject: Re: [ntpwg] questions about current NTP protocol draft


I'm honestly not EXACTLY sure what version is what in the xml domain but
I'll vote.
 
1) I don't see capepsilon in the non xml version.  Can you quote the
string or equation here?  In any case, capepsilon is root dispersion.
If the notes or context allows you to determine that the variable in
question is indeed root dispersion then EPSILON would be correct in the
latest version.
 
2) Yes, sqrt of the sum of the squares.  
 
 

Greg Dowd 
gdowd at symmetricom dot com (antispam format) 
Symmetricom, Inc. 
www.symmetricom.com 
"The current implementation is non-obvious and may need to be improved."




 


________________________________

From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org
[mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org] On Behalf
Of Plunkett, Timothy R CIV NSWCDD, W13
Sent: Wednesday, April 04, 2007 2:02 PM
To: ntpwg at lists.ntp.isc.org
Subject: [ntpwg] questions about current NTP protocol draft



I am currently editing the XML version of the latest NTPv4 protocol
draft.  I noticed the following two things that I was hoping that
someone could clarify for me.

1.  Should the equation for capepsilon in section 10 be for a variable
called EPSILON instead?  I know that there has been an effort to remove
all the "cap" variables and I think this is a leftover.

2.  Should the equation for PSI in section 11.2.3 have a square root
function in it?  Seems like PSI should equal SQRT [(PSI_s)^2 +
(PSI_p)^2].  If preferred I could also raise it to the one half power as
well.

Thanks, 

Tim 

*********************************************************** 
Timothy R. Plunkett 
NSWC-DD, Code W13 
Building 1500 Room A122A 
17214 AVENUE B SUITE 100 
DAHLGREN, VA 22448-5147 

(540) 653 - 1090 (phone at Dahlgren) 
(540) 653 - 8673 (fax at Dahlgren) 



From GDowd at symmetricom.com  Wed Apr  4 15:56:47 2007
From: GDowd at symmetricom.com (Greg Dowd)
Date: Wed, 4 Apr 2007 15:56:47 -0700
Subject: [ntpwg] questions about current NTP protocol draft
In-Reply-To: <88ED8F2A922CE84986619E21B45CD4DA02349D96@NAEAMILLEX06VA.nadsusea.nads.navy.mil>
Message-ID: <4B73D9EBE2327E40B4D69822D7C45802068D08B1@sjmail1.symmetricom.com>

I understand the question.  You should replace capepsilon with epsilon.
I was focused on the why.  I should have clearly stated the what first.


                     i=n-1
                     ---     epsilon_i
      epsilon    =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0


Greg Dowd
gdowd at symmetricom dot com (antispam format)
Symmetricom, Inc. 
www.symmetricom.com
"The current implementation is non-obvious and may need to be improved."




-----Original Message-----
From: Plunkett, Timothy R CIV NSWCDD, W13
[mailto:timothy.plunkett at navy.mil] 
Sent: Wednesday, April 04, 2007 3:17 PM
To: Greg Dowd; Odonoghue, Karen F CIV NSWCDD, W13;
ntpwg at lists.ntp.isc.org
Subject: RE: [ntpwg] questions about current NTP protocol draft

Greg,

But what I am saying is that 


                     i=n-1
                     ---     epsilon_i
      capepsilon =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0

should be changed to be instead


                     i=n-1
                     ---     epsilon_i
         EPSILON =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0
or maybe the capepsilon should only be epsilon, but I am sure that
"capepsilon" should no longer be there.  There are no other variables in
the document that start with "cap".  They have been replaced with an ALL
CAPS version of them.

Tim


***********************************************************
Timothy R. Plunkett
NSWC-DD, Code W13
Building 1500 Room A122A
17214 AVENUE B SUITE 100
DAHLGREN, VA 22448-5147

(540) 653 - 1090 (phone at Dahlgren)
(540) 653 - 8673 (fax at Dahlgren)


-----Original Message-----
From: Greg Dowd [mailto:GDowd at symmetricom.com]
Sent: Wednesday, April 04, 2007 18:10
To: Odonoghue, Karen F CIV NSWCDD, W13; Plunkett, Timothy R CIV NSWCDD,
W13; ntpwg at lists.ntp.isc.org
Subject: RE: [ntpwg] questions about current NTP protocol draft

That's peer dispersion epsilon, as opposed to root dispersion
capepsilon.    The text in the graphic shows the correct epsilon.
 

Greg Dowd
gdowd at symmetricom dot com (antispam format) Symmetricom, Inc. 
www.symmetricom.com
"The current implementation is non-obvious and may need to be improved."




 

________________________________

From: Odonoghue, Karen F CIV NSWCDD, W13
[mailto:karen.odonoghue at navy.mil]
Sent: Wednesday, April 04, 2007 2:56 PM
To: Greg Dowd; Plunkett, Timothy R CIV NSWCDD, W13;
ntpwg at lists.ntp.isc.org
Subject: RE: [ntpwg] questions about current NTP protocol draft


I asked Tim to check the reformatted xml to make sure nothing got
corrupted in the process. 
The equation in question is: 
 
Karen
 
   In the next step the shift register stages are copied to a temporary
   list and the list sorted by increasing delta.  Let i index the stages
   starting with the lowest delta.  If the first tuple epoch t_0 is not
   later than the last valid sample epoch p.t, the routine exits without
   affecting the current peer variables.  Otherwise, let epsilon_i be
   the dispersion of the ith entry, then
                     i=n-1
                     ---     epsilon_i
      capepsilon =    \     ----------
                      /        (i+1)
                     ---     2
                     i=0

   is the peer dispersion p.disp.  Note the overload of epsilon, whether
   input to the clock filter or output, the meaning should be clear from
   context.


________________________________

From: ntpwg-bounces+karen.odonoghue=navy.mil at lists.ntp.isc.org on behalf
of Greg Dowd
Sent: Wed 4/4/2007 4:22 PM
To: Plunkett, Timothy R CIV NSWCDD, W13; ntpwg at lists.ntp.isc.org
Subject: Re: [ntpwg] questions about current NTP protocol draft


I'm honestly not EXACTLY sure what version is what in the xml domain but
I'll vote.
 
1) I don't see capepsilon in the non xml version.  Can you quote the
string or equation here?  In any case, capepsilon is root dispersion.
If the notes or context allows you to determine that the variable in
question is indeed root dispersion then EPSILON would be correct in the
latest version.
 
2) Yes, sqrt of the sum of the squares.  
 
 

Greg Dowd 
gdowd at symmetricom dot com (antispam format) 
Symmetricom, Inc. 
www.symmetricom.com 
"The current implementation is non-obvious and may need to be improved."




 


________________________________

From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org
[mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org] On Behalf
Of Plunkett, Timothy R CIV NSWCDD, W13
Sent: Wednesday, April 04, 2007 2:02 PM
To: ntpwg at lists.ntp.isc.org
Subject: [ntpwg] questions about current NTP protocol draft



I am currently editing the XML version of the latest NTPv4 protocol
draft.  I noticed the following two things that I was hoping that
someone could clarify for me.

1.  Should the equation for capepsilon in section 10 be for a variable
called EPSILON instead?  I know that there has been an effort to remove
all the "cap" variables and I think this is a leftover.

2.  Should the equation for PSI in section 11.2.3 have a square root
function in it?  Seems like PSI should equal SQRT [(PSI_s)^2 +
(PSI_p)^2].  If preferred I could also raise it to the one half power as
well.

Thanks, 

Tim 

*********************************************************** 
Timothy R. Plunkett 
NSWC-DD, Code W13 
Building 1500 Room A122A 
17214 AVENUE B SUITE 100 
DAHLGREN, VA 22448-5147 

(540) 653 - 1090 (phone at Dahlgren) 
(540) 653 - 8673 (fax at Dahlgren) 



From mills at udel.edu  Thu Apr  5 20:23:05 2007
From: mills at udel.edu (David L. Mills)
Date: Fri, 06 Apr 2007 04:23:05 +0100
Subject: [ntpwg] questions about current NTP protocol draft
In-Reply-To: <88ED8F2A922CE84986619E21B45CD4DA02349D94@NAEAMILLEX06VA.nadsusea.nads.navy.mil>
References: <88ED8F2A922CE84986619E21B45CD4DA02349D94@NAEAMILLEX06VA.nadsusea.nads.navy.mil>
Message-ID: <4615BD19.7070000@udel.edu>

Tim,

The XML I posted on the NTP Project Page had no capepsilon. Neither does 
the RFC that was made from it. What's going on?

Dave

Plunkett, Timothy R CIV NSWCDD, W13 wrote:

> I am currently editing the XML version of the latest NTPv4 protocol
> draft. I noticed the following two things that I was hoping that
> someone could clarify for me.
>
> 1. Should the equation for capepsilon in section 10 be for a variable
> called EPSILON instead? I know that there has been an effort to remove
> all the "cap" variables and I think this is a leftover.
>
> 2. Should the equation for PSI in section 11.2.3 have a square root
> function in it? Seems like PSI should equal SQRT [(PSI_s)^2 +
> (PSI_p)^2]. If preferred I could also raise it to the one half power as
> well.
>
> Thanks,
>
> Tim
>
> ***********************************************************
> Timothy R. Plunkett
> NSWC-DD, Code W13
> Building 1500 Room A122A
> 17214 AVENUE B SUITE 100
> DAHLGREN, VA 22448-5147
>
> (540) 653 - 1090 (phone at Dahlgren)
> (540) 653 - 8673 (fax at Dahlgren)
>
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> ntpwg mailing list
> ntpwg at lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/ntpwg


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070406/9a5e64da/attachment.html 

From mills at udel.edu  Thu Apr  5 20:26:53 2007
From: mills at udel.edu (David L. Mills)
Date: Fri, 06 Apr 2007 04:26:53 +0100
Subject: [ntpwg] questions about current NTP protocol draft
In-Reply-To: <4B73D9EBE2327E40B4D69822D7C45802068D08AF@sjmail1.symmetricom.com>
References: <4B73D9EBE2327E40B4D69822D7C45802068D08AF@sjmail1.symmetricom.com>
Message-ID: <4615BDFD.2080007@udel.edu>

Greg,

Ths history as I know it started from the XML document I posted on the 
NTP project page. Harlan took that and produced the RFC. I purposely 
removed the XML here so as not to confuse the markup. So far as I know, 
Harlan has the only XML copy. I'm sure he would be hapy to give it up; 
maybe he already has. However, if you see a capepsilon, you have the 
wrong document.

Dave

Greg Dowd wrote:

> I'm honestly not EXACTLY sure what version is what in the xml domain but
> I'll vote.
>
> 1) I don't see capepsilon in the non xml version. Can you quote the
> string or equation here? In any case, capepsilon is root dispersion.
> If the notes or context allows you to determine that the variable in
> question is indeed root dispersion then EPSILON would be correct in the
> latest version.
>
> 2) Yes, sqrt of the sum of the squares.
>
>
>
> Greg Dowd
> gdowd at symmetricom dot com (antispam format)
> Symmetricom, Inc.
> www.symmetricom.com
> "The current implementation is non-obvious and may need to be improved."
>
>
>
>
>
>
>
> ________________________________
>
> From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org
> [mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org] On Behalf
> Of Plunkett, Timothy R CIV NSWCDD, W13
> Sent: Wednesday, April 04, 2007 2:02 PM
> To: ntpwg at lists.ntp.isc.org
> Subject: [ntpwg] questions about current NTP protocol draft
>
>
>
> I am currently editing the XML version of the latest NTPv4 protocol
> draft. I noticed the following two things that I was hoping that
> someone could clarify for me.
>
> 1. Should the equation for capepsilon in section 10 be for a variable
> called EPSILON instead? I know that there has been an effort to remove
> all the "cap" variables and I think this is a leftover.
>
> 2. Should the equation for PSI in section 11.2.3 have a square root
> function in it? Seems like PSI should equal SQRT [(PSI_s)^2 +
> (PSI_p)^2]. If preferred I could also raise it to the one half power as
> well.
>
> Thanks,
>
> Tim
>
> ***********************************************************
> Timothy R. Plunkett
> NSWC-DD, Code W13
> Building 1500 Room A122A
> 17214 AVENUE B SUITE 100
> DAHLGREN, VA 22448-5147
>
> (540) 653 - 1090 (phone at Dahlgren)
> (540) 653 - 8673 (fax at Dahlgren)
>
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> ntpwg mailing list
> ntpwg at lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/ntpwg


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070406/fa26a785/attachment.html 

From mills at udel.edu  Thu Apr  5 20:29:41 2007
From: mills at udel.edu (David L. Mills)
Date: Fri, 06 Apr 2007 04:29:41 +0100
Subject: [ntpwg] questions about current NTP protocol draft
In-Reply-To: <4B73D9EBE2327E40B4D69822D7C45802068D08B0@sjmail1.symmetricom.com>
References: <4B73D9EBE2327E40B4D69822D7C45802068D08B0@sjmail1.symmetricom.com>
Message-ID: <4615BEA5.5060800@udel.edu>

Guys,

It may be useful to compare the equation variables with the tables of 
variables. You should not see capepsilon anywhere. I continue to believe 
this is the wrong XML document.

Dave

Greg Dowd wrote:

> That's peer dispersion epsilon, as opposed to root dispersion
> capepsilon. The text in the graphic shows the correct epsilon.
>
>
> Greg Dowd
> gdowd at symmetricom dot com (antispam format)
> Symmetricom, Inc.
> www.symmetricom.com
> "The current implementation is non-obvious and may need to be improved."
>
>
>
>
>
>
> ________________________________
>
> From: Odonoghue, Karen F CIV NSWCDD, W13
> [mailto:karen.odonoghue at navy.mil]
> Sent: Wednesday, April 04, 2007 2:56 PM
> To: Greg Dowd; Plunkett, Timothy R CIV NSWCDD, W13;
> ntpwg at lists.ntp.isc.org
> Subject: RE: [ntpwg] questions about current NTP protocol draft
>
>
> I asked Tim to check the reformatted xml to make sure nothing got
> corrupted in the process.
> The equation in question is:
>
> Karen
>
> In the next step the shift register stages are copied to a temporary
> list and the list sorted by increasing delta. Let i index the stages
> starting with the lowest delta. If the first tuple epoch t_0 is not
> later than the last valid sample epoch p.t, the routine exits without
> affecting the current peer variables. Otherwise, let epsilon_i be
> the dispersion of the ith entry, then
> i=n-1
> --- epsilon_i
> capepsilon = \ ----------
> / (i+1)
> --- 2
> i=0
>
> is the peer dispersion p.disp. Note the overload of epsilon, whether
> input to the clock filter or output, the meaning should be clear from
> context.
>
>
> ________________________________
>
> From: ntpwg-bounces+karen.odonoghue=navy.mil at lists.ntp.isc.org on behalf
> of Greg Dowd
> Sent: Wed 4/4/2007 4:22 PM
> To: Plunkett, Timothy R CIV NSWCDD, W13; ntpwg at lists.ntp.isc.org
> Subject: Re: [ntpwg] questions about current NTP protocol draft
>
>
> I'm honestly not EXACTLY sure what version is what in the xml domain but
> I'll vote.
>
> 1) I don't see capepsilon in the non xml version. Can you quote the
> string or equation here? In any case, capepsilon is root dispersion.
> If the notes or context allows you to determine that the variable in
> question is indeed root dispersion then EPSILON would be correct in the
> latest version.
>
> 2) Yes, sqrt of the sum of the squares.
>
>
>
> Greg Dowd
> gdowd at symmetricom dot com (antispam format)
> Symmetricom, Inc.
> www.symmetricom.com
> "The current implementation is non-obvious and may need to be improved."
>
>
>
>
>
>
>
> ________________________________
>
> From: ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org
> [mailto:ntpwg-bounces+gdowd=symmetricom.com at lists.ntp.isc.org] On Behalf
> Of Plunkett, Timothy R CIV NSWCDD, W13
> Sent: Wednesday, April 04, 2007 2:02 PM
> To: ntpwg at lists.ntp.isc.org
> Subject: [ntpwg] questions about current NTP protocol draft
>
>
>
> I am currently editing the XML version of the latest NTPv4 protocol
> draft. I noticed the following two things that I was hoping that
> someone could clarify for me.
>
> 1. Should the equation for capepsilon in section 10 be for a variable
> called EPSILON instead? I know that there has been an effort to remove
> all the "cap" variables and I think this is a leftover.
>
> 2. Should the equation for PSI in section 11.2.3 have a square root
> function in it? Seems like PSI should equal SQRT [(PSI_s)^2 +
> (PSI_p)^2]. If preferred I could also raise it to the one half power as
> well.
>
> Thanks,
>
> Tim
>
> ***********************************************************
> Timothy R. Plunkett
> NSWC-DD, Code W13
> Building 1500 Room A122A
> 17214 AVENUE B SUITE 100
> DAHLGREN, VA 22448-5147
>
> (540) 653 - 1090 (phone at Dahlgren)
> (540) 653 - 8673 (fax at Dahlgren)
>
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> ntpwg mailing list
> ntpwg at lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/ntpwg


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070406/7e2b5d87/attachment.html 

From Helen.Y.Chen at motorola.com  Wed Apr 25 15:05:27 2007
From: Helen.Y.Chen at motorola.com (Chen Helen-A12587)
Date: Wed, 25 Apr 2007 18:05:27 -0400
Subject: [ntpwg] NTPv4 Client / NTPv3 Server
Message-ID: <655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com>

Hi Dave,
 
Does the NTPv4 client send an NTPv4 request regardless of the server's
NTP version? Will the NTPv3 server handle an NTPv4 request properly from
a client (due to ALL NTPv4 fields being optional) or must the
NTPv4-capable client send an NTPv3 request to the server?  If the
latter, how does the client know which version to send?
 
Thanks,
Helen 
 
 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070425/5f7baf9d/attachment.html 

From stenn at ntp.isc.org  Wed Apr 25 15:40:10 2007
From: stenn at ntp.isc.org (Harlan Stenn)
Date: Wed, 25 Apr 2007 22:40:10 +0000
Subject: [ntpwg] NTPv4 Client / NTPv3 Server
In-Reply-To: Message from "Chen Helen-A12587" <Helen.Y.Chen@motorola.com> 
	of "Wed, 25 Apr 2007 18:05:27 -0400."
	<655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com>
Message-ID: <20070425224015.C3A9239D08@ntp1.ntp.isc.org>

Helen,

The one who talks first send the highest version it can handle.

The other side responds with whichever is less, the sender's version or
the "local" version.

>From then people behave according to that version.

H

From Helen.Y.Chen at motorola.com  Wed Apr 25 15:54:34 2007
From: Helen.Y.Chen at motorola.com (Chen Helen-A12587)
Date: Wed, 25 Apr 2007 18:54:34 -0400
Subject: [ntpwg] NTPv4 Client / NTPv3 Server
In-Reply-To: <20070425224015.8979739CFD@ntp1.ntp.isc.org>
References: Message from "Chen Helen-A12587" <Helen.Y.Chen@motorola.com> of
	"Wed, 25 Apr 2007 18:05:27 -0400."
	<655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com>
	<20070425224015.8979739CFD@ntp1.ntp.isc.org>
Message-ID: <655403F33031E14DB1EDF1422873047501A5E22A@de01exm68.ds.mot.com>

Harlan,

Thanks. Isn't it always the client who talks first though, since it's
requesting the time or authentication?

Helen 

-----Original Message-----
From: Harlan Stenn [mailto:stenn at ntp.isc.org] 
Sent: Wednesday, April 25, 2007 5:40 PM
To: Chen Helen-A12587
Cc: David L. Mills; ntpwg at ntp.isc.org
Subject: Re: [ntpwg] NTPv4 Client / NTPv3 Server 

Helen,

The one who talks first send the highest version it can handle.

The other side responds with whichever is less, the sender's version or
the "local" version.

>From then people behave according to that version.

H

From tglassey at earthlink.net  Wed Apr 25 16:52:56 2007
From: tglassey at earthlink.net (todd glassey)
Date: Wed, 25 Apr 2007 16:52:56 -0700
Subject: [ntpwg] NTPv4 Client / NTPv3 Server
References: Message from "Chen Helen-A12587" <Helen.Y.Chen@motorola.com>
	of"Wed, 25 Apr 2007 18:05:27
	-0400."<655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com><20070425224015.8979739CFD@ntp1.ntp.isc.org>
	<655403F33031E14DB1EDF1422873047501A5E22A@de01exm68.ds.mot.com>
Message-ID: <002901c78794$da5e4830$174f7d40@home.glassey.com>

Only in Unicast Events. 

Todd 
----- Original Message ----- 
From: "Chen Helen-A12587" <Helen.Y.Chen at motorola.com>
To: "Harlan Stenn" <stenn at ntp.isc.org>
Cc: <ntpwg at ntp.isc.org>
Sent: Wednesday, April 25, 2007 3:54 PM
Subject: Re: [ntpwg] NTPv4 Client / NTPv3 Server


> Harlan,
> 
> Thanks. Isn't it always the client who talks first though, since it's
> requesting the time or authentication?
> 
> Helen 
> 
> -----Original Message-----
> From: Harlan Stenn [mailto:stenn at ntp.isc.org] 
> Sent: Wednesday, April 25, 2007 5:40 PM
> To: Chen Helen-A12587
> Cc: David L. Mills; ntpwg at ntp.isc.org
> Subject: Re: [ntpwg] NTPv4 Client / NTPv3 Server 
> 
> Helen,
> 
> The one who talks first send the highest version it can handle.
> 
> The other side responds with whichever is less, the sender's version or
> the "local" version.
> 
>>From then people behave according to that version.
> 
> H
> _______________________________________________
> ntpwg mailing list
> ntpwg at lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/ntpwg

From mills at udel.edu  Thu Apr 26 10:06:19 2007
From: mills at udel.edu (David L. Mills)
Date: Thu, 26 Apr 2007 17:06:19 +0000
Subject: [ntpwg] NTPv4 Client / NTPv3 Server
In-Reply-To: <655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com>
References: <655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com>
Message-ID: <4630DC0B.8060901@udel.edu>

Helen,

The ntp client sends in the configured version from 1 through the client 
version. The ntp server responds with the version specified by the 
client, but not greater than the server version. Thus, a NTPv4 server 
responds to any version up to version 4, while a NTPv3 server responds 
to any version up to version 3.

Dave

Chen Helen-A12587 wrote:

> Hi Dave,
>  
> Does the NTPv4 client send an NTPv4 request regardless of the server's
> NTP version? Will the NTPv3 server handle an NTPv4 request properly from
> a client (due to ALL NTPv4 fields being optional) or must the
> NTPv4-capable client send an NTPv3 request to the server?  If the
> latter, how does the client know which version to send?
>  
> Thanks,
> Helen 
>  
>  
>  
> 

From mills at udel.edu  Thu Apr 26 10:28:19 2007
From: mills at udel.edu (David L. Mills)
Date: Thu, 26 Apr 2007 17:28:19 +0000
Subject: [ntpwg] NTPv4 Client / NTPv3 Server
In-Reply-To: <655403F33031E14DB1EDF1422873047501A5E22A@de01exm68.ds.mot.com>
References: Message from "Chen Helen-A12587" <Helen.Y.Chen@motorola.com> of
	"Wed, 25 Apr 2007 18:05:27 -0400."
	<655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com>
	<20070425224015.8979739CFD@ntp1.ntp.isc.org>
	<655403F33031E14DB1EDF1422873047501A5E22A@de01exm68.ds.mot.com>
Message-ID: <4630E133.2030609@udel.edu>

Helen,

The client ordinarily sends its compiled version; however, an option can 
be specified to send in any other version.

Chen Helen-A12587 wrote:

> Harlan,
> 
> Thanks. Isn't it always the client who talks first though, since it's
> requesting the time or authentication?
> 
> Helen 
> 
> -----Original Message-----
> From: Harlan Stenn [mailto:stenn at ntp.isc.org] 
> Sent: Wednesday, April 25, 2007 5:40 PM
> To: Chen Helen-A12587
> Cc: David L. Mills; ntpwg at ntp.isc.org
> Subject: Re: [ntpwg] NTPv4 Client / NTPv3 Server 
> 
> Helen,
> 
> The one who talks first send the highest version it can handle.
> 
> The other side responds with whichever is less, the sender's version or
> the "local" version.
> 
> From then people behave according to that version.
> 
> H

From Helen.Y.Chen at motorola.com  Thu Apr 26 12:52:21 2007
From: Helen.Y.Chen at motorola.com (Chen Helen-A12587)
Date: Thu, 26 Apr 2007 15:52:21 -0400
Subject: [ntpwg] [ntp:hackers] Autokey Identity Scheme Selection
References: Your message of "Wed, 27 Sep 2006 17:07:58 EDT."
	<655403F33031E14DB1EDF14228730475012721F8@de01exm68.ds.mot.com>
	<E1GTIMA-0005qo-U3@stasis.kostecke.net>
	<655403F33031E14DB1EDF1422873047501A5E228@de01exm68.ds.mot.com>
	<4630E0D7.1090807@udel.edu> 
Message-ID: <655403F33031E14DB1EDF1422873047501A5E57E@de01exm68.ds.mot.com>

Dave,

Another question to add:
>From what I understand, the group parameters are distributed by the TA
(or stratum 1 server, if this is the TA as well) to both the servers AND
the clients in the group. My impresssion is that the clients are sent
the group parameters EXCLUDING the group key. Is this true? If this is
true, is it possible for an intermediate server to distribute the
non-group key parametrs to the clients? Is it possible for the
intermediate server to separate out the group key from the other
parameters, in order to send the correct parameters to the client?

Thanks,
Helen



-----Original Message-----
From: Chen Helen-A12587 
Sent: Thursday, April 26, 2007 12:49 PM
To: 'David L. Mills'
Cc: Steve Kostecke
Subject: RE: [ntp:hackers] Autokey Identity Scheme Selection

Dave,

Are your statements applicable to the IFF scheme as well? In other
words, can the certificate for the IFF scheme be generated by the NTP
tools on the stratum 1 server (at the top of the certificate trail)?

Is changing the group parameters for the scheme something done outside
of the protocol, where someone would have to re-generate the parameters
again and restart the protocol by restarting the ntpd?

Thanks,
Helen 

-----Original Message-----
From: David L. Mills [mailto:mills at udel.edu]
Sent: Thursday, April 26, 2007 12:27 PM
To: Chen Helen-A12587
Cc: Steve Kostecke
Subject: Re: [ntp:hackers] Autokey Identity Scheme Selection

Helen,

Ordinarily the certificate is generated by the NTP tools. In principle,
the certificate could be generated by some other entity using the
OpenSSL library or another library.

The parameters for generating the public/private key pairs must be the
same for all servers and clients, but could in principle be other than
those assumed by default.

The reference implementation includes a utility supporting all modes
specified in stime.pdf.

Dave

Chen Helen-A12587 wrote:

> Steve, Dave,
> 
> 1) Stime.pdf states that IFF scheme should be used when the 
> certificate is generated by a commercial service. What if the 
> certificate is NOT generated by a commercial service, e.g. if it's 
> generated by the root time server? Would it still be OK to use the IFF

> scheme? Should the GQ scheme be used instead (as suggested in
> stime.pdf) when the parameters are generated by the ntp routines 
> included in the reference implementation?
> 
> 2) Is the changing of the group parameters a matter of local policy? 
> Is this done (1) automatically by the software or (2) manually by 
> executing ntp-keygen again, and restarting the ntpd?
> 
> 3) I assume that the reference implementation provides a different 
> 'ntpkey' routine for each of the identity schemes it supports. Is this

> true?
> 
> Thanks,
> Helen
> 
> -----Original Message-----
> From: Steve Kostecke [mailto:steve at kostecke.net] On Behalf Of Steve 
> Kostecke
> Sent: Friday, September 29, 2006 8:25 AM
> To: Chen Helen-A12587
> Subject: Re: [ntp:hackers] Autokey Identity Scheme Selection
> 
> "Chen Helen-A12587" said:
> 
> 
>>The documentation in ntp.isc.org mentions that the reference 
>>implementation includes 3 identity schemes (IFF, GQ and MV). How does 
>>one configure the client and server to use a particular identity
> 
> scheme?
> 
> You activate Identity Schemes by placing the Identity Scheme 
> "Parameters" in the ntpkeys directories on all members of the NTP 
> Trust Group. This is in addition to the other configuration steps 
> outlined in http://ntp.isc.org/Support/ConfiguringAutokey
> 
> You first set-up autokey on the Trust Group server as outlined in 
> http://ntp.isc.org/bin/view/Support/ConfiguringAutokey#Section_6.6.1.
> 
> When you set-up the Trust Group server you generate the "Parameters" 
> for the Identity Scheme(s) you wish to use in that Trust Group. The 
> server loads all of the Identity Scheme "Parameters" located in the 
> ntpkeys directory and activates those Identity Schemes.
> 
> Then you set-up autokey on the Trust Group Clients as outlined in 
> http://ntp.isc.org/bin/view/Support/ConfiguringAutokey#Section_6.6.2.
> 
> Part of the Trust Group client set-up process involves acquiring the 
> Group/Client Keys for as outlined in 
>
http://ntp.isc.org/bin/view/Support/ConfiguringAutokey#Section_6.6.2.4.
> 
> The Trust Group clients will attempt to use the Identity Scheme 
> "Parameters" found in their ntpkeys directories.
> 
> It is generally recommended that you start with IFF.
> 
> --
> Steve Kostecke <kostecke at ntp.isc.org>
> NTP Public Services Project http://ntp.isc.org/ Public Key at 
> http://ntp.isc.org/Users/SteveKostecke

From mills at udel.edu  Thu Apr 26 19:40:39 2007
From: mills at udel.edu (David L. Mills)
Date: Fri, 27 Apr 2007 02:40:39 +0000
Subject: [ntpwg] [ntp:hackers] Autokey Identity Scheme Selection
In-Reply-To: <655403F33031E14DB1EDF1422873047501A5E57E@de01exm68.ds.mot.com>
References: Your message of "Wed, 27 Sep 2006 17:07:58 EDT."
	<655403F33031E14DB1EDF14228730475012721F8@de01exm68.ds.mot.com>
	<E1GTIMA-0005qo-U3@stasis.kostecke.net>
	<655403F33031E14DB1EDF1422873047501A5E228@de01exm68.ds.mot.com>
	<4630E0D7.1090807@udel.edu>
	<655403F33031E14DB1EDF1422873047501A5E57E@de01exm68.ds.mot.com>
Message-ID: <463162A7.5030406@udel.edu>

Helen,

Depends upon the scheme. See 
http://www.eecis.udel.edu/~mills/ident.html. The intent is that the 
servers and clients get their keys from the TA via a secure web page, 
however that is done.

Dave

Chen Helen-A12587 wrote:

> Dave,
>
> Another question to add:
> From what I understand, the group parameters are distributed by the TA
> (or stratum 1 server, if this is the TA as well) to both the servers AND
> the clients in the group. My impresssion is that the clients are sent
> the group parameters EXCLUDING the group key. Is this true? If this is
> true, is it possible for an intermediate server to distribute the
> non-group key parametrs to the clients? Is it possible for the
> intermediate server to separate out the group key from the other
> parameters, in order to send the correct parameters to the client?
>
> Thanks,
> Helen
>
>
>
> -----Original Message-----
> From: Chen Helen-A12587
> Sent: Thursday, April 26, 2007 12:49 PM
> To: 'David L. Mills'
> Cc: Steve Kostecke
> Subject: RE: [ntp:hackers] Autokey Identity Scheme Selection
>
> Dave,
>
> Are your statements applicable to the IFF scheme as well? In other
> words, can the certificate for the IFF scheme be generated by the NTP
> tools on the stratum 1 server (at the top of the certificate trail)?
>
> Is changing the group parameters for the scheme something done outside
> of the protocol, where someone would have to re-generate the parameters
> again and restart the protocol by restarting the ntpd?
>
> Thanks,
> Helen
>
> -----Original Message-----
> From: David L. Mills [mailto:mills at udel.edu]
> Sent: Thursday, April 26, 2007 12:27 PM
> To: Chen Helen-A12587
> Cc: Steve Kostecke
> Subject: Re: [ntp:hackers] Autokey Identity Scheme Selection
>
> Helen,
>
> Ordinarily the certificate is generated by the NTP tools. In principle,
> the certificate could be generated by some other entity using the
> OpenSSL library or another library.
>
> The parameters for generating the public/private key pairs must be the
> same for all servers and clients, but could in principle be other than
> those assumed by default.
>
> The reference implementation includes a utility supporting all modes
> specified in stime.pdf.
>
> Dave
>
> Chen Helen-A12587 wrote:
>
>> Steve, Dave,
>>
>> 1) Stime.pdf states that IFF scheme should be used when the
>> certificate is generated by a commercial service. What if the
>> certificate is NOT generated by a commercial service, e.g. if it's
>> generated by the root time server? Would it still be OK to use the IFF
>
>
>> scheme? Should the GQ scheme be used instead (as suggested in
>> stime.pdf) when the parameters are generated by the ntp routines
>> included in the reference implementation?
>>
>> 2) Is the changing of the group parameters a matter of local policy?
>> Is this done (1) automatically by the software or (2) manually by
>> executing ntp-keygen again, and restarting the ntpd?
>>
>> 3) I assume that the reference implementation provides a different
>> 'ntpkey' routine for each of the identity schemes it supports. Is this
>
>
>> true?
>>
>> Thanks,
>> Helen
>>
>> -----Original Message-----
>> From: Steve Kostecke [mailto:steve at kostecke.net] On Behalf Of Steve
>> Kostecke
>> Sent: Friday, September 29, 2006 8:25 AM
>> To: Chen Helen-A12587
>> Subject: Re: [ntp:hackers] Autokey Identity Scheme Selection
>>
>> "Chen Helen-A12587" said:
>>
>>
>>> The documentation in ntp.isc.org mentions that the reference
>>> implementation includes 3 identity schemes (IFF, GQ and MV). How does
>>> one configure the client and server to use a particular identity
>>
>> scheme?
>>
>> You activate Identity Schemes by placing the Identity Scheme
>> "Parameters" in the ntpkeys directories on all members of the NTP
>> Trust Group. This is in addition to the other configuration steps
>> outlined in http://ntp.isc.org/Support/ConfiguringAutokey
>>
>> You first set-up autokey on the Trust Group server as outlined in
>> http://ntp.isc.org/bin/view/Support/ConfiguringAutokey#Section_6.6.1.
>>
>> When you set-up the Trust Group server you generate the "Parameters"
>> for the Identity Scheme(s) you wish to use in that Trust Group. The
>> server loads all of the Identity Scheme "Parameters" located in the
>> ntpkeys directory and activates those Identity Schemes.
>>
>> Then you set-up autokey on the Trust Group Clients as outlined in
>> http://ntp.isc.org/bin/view/Support/ConfiguringAutokey#Section_6.6.2.
>>
>> Part of the Trust Group client set-up process involves acquiring the
>> Group/Client Keys for as outlined in
>>
> http://ntp.isc.org/bin/view/Support/ConfiguringAutokey#Section_6.6.2.4.
>
>> The Trust Group clients will attempt to use the Identity Scheme
>> "Parameters" found in their ntpkeys directories.
>>
>> It is generally recommended that you start with IFF.
>>
>> --
>> Steve Kostecke <kostecke at ntp.isc.org>
>> NTP Public Services Project http://ntp.isc.org/ Public Key at
>> http://ntp.isc.org/Users/SteveKostecke
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.isc.org/pipermail/ntpwg/attachments/20070427/e13a9ca7/attachment.html 

From kostecke at ntp.isc.org  Thu Apr 26 19:51:50 2007
From: kostecke at ntp.isc.org (Steve Kostecke)
Date: Thu, 26 Apr 2007 22:51:50 -0400
Subject: [ntpwg] [ntp:hackers] Autokey Identity Scheme Selection
In-Reply-To: Your message of "Fri, 27 Apr 2007 02:40:39 -0000."
	<463162A7.5030406@udel.edu> 
Message-ID: <E1HhGYo-000744-UI@stasis.kostecke.net>

"David L. Mills" said:

>Chen Helen-A12587 wrote:
>
>> Is it possible for the intermediate server to separate out the
>> group key from the other parameters, in order to send the correct
>> parameters to the client?
>
>Depends upon the scheme. See
>http://www.eecis.udel.edu/~mills/ident.html. The intent is that the
>servers and clients get their keys from the TA via a secure web page,
>however that is done.

Back in 2004 I wrote an IFF Client Key request page which is located at
https://ntp.isc.org/crypto.php You can use it to retrieve a working IFF
Client Key for the time server at ntp2.isc.org.

-- 
Steve Kostecke <kostecke at ntp.isc.org>
NTP Public Services Project http://ntp.isc.org/
Public Key at http://ntp.isc.org/Users/SteveKostecke

From mayer at ntp.isc.org  Sun Apr 29 10:00:27 2007
From: mayer at ntp.isc.org (Danny Mayer)
Date: Sun, 29 Apr 2007 13:00:27 -0400
Subject: [ntpwg] NTPv4 Client / NTPv3 Server
In-Reply-To: <655403F33031E14DB1EDF1422873047501A5E22A@de01exm68.ds.mot.com>
References: Message from "Chen Helen-A12587" <Helen.Y.Chen@motorola.com>
	of	"Wed, 25 Apr 2007 18:05:27
	-0400."	<655403F33031E14DB1EDF1422873047501A5E208@de01exm68.ds.mot.com>	<20070425224015.8979739CFD@ntp1.ntp.isc.org>
	<655403F33031E14DB1EDF1422873047501A5E22A@de01exm68.ds.mot.com>
Message-ID: <4634CF2B.70703@ntp.isc.org>

Chen Helen-A12587 wrote:
> Harlan,
> 
> Thanks. Isn't it always the client who talks first though, since it's
> requesting the time or authentication?
> 
> Helen 
> 

Not always. If the server is a broadcast or multicast server sendimg
broadcast packets the server will talk first.

Danny


