-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Ask Bjørn Hansen wrote:
>
> On Jul 2, 2007, at 2:13, M Graff wrote:
>> I don't see how the compression thing really factors in all that much?
>
> Until we can (more easily) make the DNS servers just not send the
> authoritative and additional sections then we save a few bytes by having
> the nameservers be
>
> {a..e}.ntpns.org rather than
>
> ns1.some-domain.net
> foo.another-domain.some.tld
> etc
Sure, but there's no guarantee that even if you do modify your DNS
servers to not send some sections (carefully done, to make certain you
don't violate standards, I assume) there's no guarantee a recursive
server somewhere won't send them.
Very few NTP clients will talk to your DNS servers. They probably talk
to local ones, which then talk to yours, and no matter how carefully you
construct your DNS responses, those caching name servers will do as they
please.
Besides, there is something to be said about having the hosts in NS
records be in different domains. If you put them all in one, say:
ntp-pool.org NS a.ntpns.org.
NS b.ntpns.org.
NS c.ntpns.org.
you have then created a requirement that ntpns.org's name servers be up
in order to resolve ntp-pool.org as well, in many cases. Sure, glue
will help somewhat, but only somewhat.
It might be better to have diversity here and have a NS record in the
.org namespace, .com namespace, and perhaps another TLD.
Packet size really isn't that much of an issue. You can pack at least
22 A records in a typical response of 512 bytes max. Far more if you
assume EDNS0 is in play, which many DNS servers use by default.
- --Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (MingW32)
iD8DBQFGiNZzuzMQWQwZDN0RAvoPAJkByq40Vi18F2NNkLAZbEh2Ba7VgACfV7PG
Yr75jdeXi3y/tuELrAyUJcM=
=X8bj
-----END PGP SIGNATURE-----
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
