[ntp:questions] Re: Tinker and tos configuration commands....
David L. Mills
mills at udel.edu
Tue Feb 8 04:18:39 UTC 2005
I have the ideal solution to your concerns. A task force on NTPv4
specificaiton is lighting up. As with rfc1305, as the specification
process matures, little changes will no doubt be made to harmonize the
specifiation and implemenation. At that time I/we will address the
open/closed interval issue. Meanwhile, and as a practical matter, the
difference really doesn't matter.
I have nothing against the questions list, but I do get rude messages
when I forget to strike out the questions reply. As you know, and with
due respect, I would prefer all the discussions to take place in the
The maxdist is very simple. As you know, the dispersion and therefore
the distance comes down as replies are received. When the distance falls
below maxdist, the reply is elegible for selection.
brad at shub-internet.org wrote:
> "David L. Mills" <mills at udel.edu> wrote:
>>Where are we going with this? You point out some comment
>>inconsistencies. The code that leaves me is loaded with comments, some
>>of which are fifteen years old.
> There was a bogus message to syslog, and an error in one comment. Both
> appear to stem from the same problem -- doing cut-n-paste and then
> forgetting to go back and correct something in the newly pasted portion.
> This is easily fixed, but we're talking about sections of code that are
> clearly marked as requiring your approval before any changes are made, and
> presumably very nasty things would happen to the person who made unapproved
> I'm pretty sure that I have the technical access I need to change whatever
> code I want wherever I want in the repository, but I want to make sure that
> the code changes get filtered through the proper process. In this case,
> that means talking to you.
>> As for
>>the limits checks, we are dealing with floating doubles here; the
>>difference between inclusive and exclusive limits is something like
>>10e-9. Big deal.
> Granted, the numerical difference is small. However, the limits are the
> limits, and nothing is allowed to exceed the limits, correct? In that
> case, how do you know when you have had a number that is equal to the
> limits but did not exceed them, as compared to a number that greatly
> exceeded the limits, but was truncated?
> I know we're talking about floating point math, in which nothing is ever
> exact. However, what I'm saying here is that I think we want to give
> ourselves the best possible chance to be able to detect when we are
> absolutely within the limits, and then take other action (as appropriate)
> when we reach or exceed the limits.
>> As for maxdist, see the beginning of the
>>clock_select routine. It is not the intent to set anything at startup and
>>change later, but to set parameters that apply throughout the life of the
> I've looked at that code. I think I understood most of the comments, but
> I'm still no closer on understanding how maxdist influences this process.
>>Can we get rid of the questions thing? The purpose is obscure and
>>doesn't contribute to the discussion here.
> Questions thing? You mean you want to eliminate the news-to-mail gateway
> completely? I think there are a lot of people on the mailing list that
> would be very unhappy with that prospect.
> Mailman is already configured to not add the "[ntp:questions]" tag to the
> "Subject:" header as messages come in and are posted to the newsgroup, but
> if the messages come in that way from the client (e.g., in an e-mail reply
> to the list), it doesn't try to strip any examples of the subject prefix.
> Mailman does work hard to avoid putting additional copies of the subject
> prefix on messages as they are sent out to the mailing list recipients, but
> so far as I know it doesn't attempt to strip any extra copies of the
> subject prefix that may have been placed there by misconfigured clients.
> If you want Mailman to strip the subject prefix before posting the messages
> to the newsgroup, that's going to take some fairly serious hacking on the
> code. I'm not sure I'm qualified to do that, and I don't know if any of
> the primary Mailman developers are in a place where they would be able to
> make that kind of a hack on our behalf.
More information about the questions