djbdns: An alternative to BIND

Mon Apr 11 19:55:14 UTC 2005

there are replies to three messages here, in an attempt to sweep up the
thread and/or end it.  i can see from the tailings that a lot of you are
not only reading dv8's posts, but replying to them.  i'm trying to sort
out the part of the result that's meaningful in spite of that poison.

note that the BIND9 code base is completely forkable.  some dns pirates
added code to 9.0.0 to try to make it look at more than one set of root
name servers, which was naive on their part but they were acting in the
spirit and the letter of the BSD-style license we use.  (of course, we
thought their patch was silly, so we didn't incorporate it into the
code base we publish from.)  apparently they, and their bind9 fork, are
now lost in the weeds of prehistory.

--- william at elan.net ("william(at)elan.net") writes:

> Again, the point of all of that is that they chose to implement protocol 
> that is non-standard, but knowing that they made sure that this would only 
> be used between BIND9 programs.

no, that's not what we did at all.

> This is proprietary protocol but as long as its used only when their
> products are talking to each other, there is nothing substantially
> wrong. Well ok, what maybe wrong is that they still call it AXFR instead
> of clearly calling it something like AXFR-BIND9.

let's lay this to rest, shall we?  the people who implemented BIND9 read the
spec (RFC 1035) and pronounced it "toe-may-toe", whereas the person who
implemented "tinydns" read the same spec and pronounced it "toe-mah-toe".
the BIND9 interpretation is more strict, and therefore runs afoul of the
internet philosophy to "be liberal in what you accept".  however, since
BIND9 is compatible with BIND8 and BIND4, and with microsoft's DNS, and
with virtually every other DNS in the world except for "tinydns", various
people put some effort into clarifying RFC 1035.  they encountered endless
and may i say vitriolic opposition from the "tinydns" author, and in the
end, the IETF somewhat spinelessly decided that clarification was "too hard"
and so here we are, with incompatible implementors each thinking we're right.

> Well, Paul Vixie wrote bind

nope.  kevin dunlap and other folks at U C Berkeley wrote BIND originally.
all i did was fork the code base at 4.8.3, produce King James BIND, then
BIND 4.9 through BIND 8.1, and along the way co-founded ISC with rick
adams.  also along the way i won the "most cert advisories by a single
author" award (which noone has been willing to try to take away from me)
and stopped coding.  i'm pleased to announce that BIND9 has no code from
BIND8 or BIND4 in it, and also no code from me in it.

--- oberman at es.net ("Kevin Oberman") writes:

> Paul Vixie did NOT write the original BIND. The first BIND version
> (4.3?) was written by the CSRG at UC Berkeley by Kevin Dunlap who was on
> loan to CSRG by Digital (who also employed Paul at that time).

no, i was employed by Digital later, long after kevin dunlap had moved on.

> When Paul took over support of BIND at about 4.4,

4.8.3.

> it was a horrid mess and rapidly moving toward death.

and there were other code forks besides mine.  what distinguished my work
was that i merged in every change i could understand from every other fork.
(that's why i called it King James BIND, for you literary history buffs.)

> After some fixes and clean-up of the
> code, the first real BIND from Paul was 4.8.

4.9.

> ISC (including Paul) wrote BIND 8.

john gilmore and bob halley had a LOT to do with the creation of BIND8 btw.
(john also taught me to use CVS rather than RCS, to my great betterment,
and he wrote some early DNSSEC code, and negotiated a licensing deal between
RSADSI and ISC... he's an unsung hero in the BIND revolution.)

> BIND 9 was contracted out to Nominum

internet engines.  which later became nominum.

> and one of the stipulations was that the existing code base could not be
> used at all and another was that the team that wrote BIND 8 should not
> work on BIND 9.

actually, bob halley worked on both BIND8 and BIND9.

> For that (and other) reason, Paul did not write any of BIND 9.

yea, verily.

--- david.conrad at nominum.com (David Conrad) writes:

> However, I don't speak authoritatively (pun intended) on BIND.

and yet, for the record, i agree with everything drc said in his note today.
-- 
Paul Vixie