8.3 vs 8.4

Jim Reid jim at rfc1035.com
Mon Dec 8 18:29:32 UTC 2003


>>>>> "Mark" == Mark  <admin at asarian-host.net> writes:

    >> Yes, BIND9 uses more resources than BIND8. This simply won't
    >> matter unless someone is running a name server on tiny
    >> hardware, say an M68K box with 8MB of RAM.

    Mark> Despite your rather arrogant presumption that people do not
    Mark> need a smaller memory footprint, or faster processing of
    Mark> queries, truth is, that BIND9 is bloated, slower, and still
    Mark> rather buggy.

My "arrogant presumption" is the truth for most DNS sites. Few people
know or care about the query rates their name servers get or the size
of the server's memory/VM footprint. Even fewer of them will be
environments where the recursive query rates are high and/or the RAM
is limited. A root server operator or DNS administrator at a major ISP
like AOL may have valid concerns about BIND9's performance. The rest
of us don't quite frankly.

I fear you may be confusing need with want. Sure, I want a blindingly
fast name server that doesn't use much RAM. Who doesn't? But I don't
need one. The server I use right now works perfectly well. It isn't
even close to being a major resource drain or performance bottleneck
on the 5 year-old 300MHz Pentium II that's the platform for my mail
and DNS server.

It just took my BIND9 server 322ms to resolve the (broken) MX records
for asarian-host.net. Let's ignore network latency and assume that
BIND8 was twice as fast as BIND9. So the delay would then be 160ms.
Who can tell the difference? The best case from this contrived and
hypothetical scenario is a saving of a massive tenth of a second.

    Mark> Ever since BIND9 came out there has been an
    Mark> endless stream of patches, sub-releases, and fixes for all
    Mark> sorts of things; some of which are rather serious, like
    Mark> common recursion errors. D.  J. Bernstein tracked them all
    Mark> down, and noted that since BIND 9.0.0b2, until BIND
    Mark> 9.2.2rc1, the number of official bugs ran into, yes, ... six
    Mark> hundred seventy-two!

First of all, while Dan Bernstein knows a lot about DNS, he is not an
authority on BIND. Tracking the "official bugs" is not hard: they're
listed in the change log that's shipped with the distribution. BTW,
the CHANGES file for the current BIND8 and BIND9 releases are about
the same size, not that this is a metric of code quality. Read the
code and make a judgement on that yourself.



More information about the bind-users mailing list