[ On Tuesday, June 3, 2003 at 19:00:33 (+1000), Luke Mewburn wrote: ]
> Subject: Re: BIND 8.3.5
>
> What NIH attitude?  You're advocating removing code that we already have
> and replacing it with code that is not source (API) or binary (ABI)
> compatible with previous NetBSD code.  That is not acceptable for NetBSD.

You guys are all getting this NIH stuff completely backwards.

Strictly "NIH" has nothing to with backwards compatability of one
system's interfaces -- quite the opposite in fact.  Rather NIH has to do
with some group avoiding cross-system compatability that comes from
using portable, shared, code, simply because it was "Not Invented Here".

NetBSD is already stuck in NIH mode _because_ it is not adopting the
common, complete, shared, BIND-8 resolver library but rather is sticking
to use of code that was invented solely for use in NetBSD.  I.e. NetBSD
is effectively saying to its users that the the BIND code was not
invented by and for NetBSD so it must not be worth using.  Avoiding use
of the BIND code just because NetBSD wants to preserve backward
compatability to what is effectively a very proprietary API is the very
definition of NIH.

Personally I don't care about BIND's IRS vs. NetBSD's nsswitch because I
don't make use of either one.  I just want NetBSD to benefit from having
integrated a widely used and widely maintained DNS resolver library.

Note that what FreeBSD have done, especially in their 4-STABLE branch,
and what GLIBC have done, is _NOT_ useful as both have VERY seriously
botched the BIND-8 resolver API -- i.e. made a complete and total mess
of it.

If we're going to ditch BIND's IRS in favour of keeping the current
nsswitch exactly as-is, then PLEASE let us work with the BIND
maintainers to make IRS and nsswitch both "pluggable" (i.e. at compile
time) in the base BIND release so that our nsswitch can be substitued
into any and every new release of the BIND resolver -- i.e. so that new
releases of the BIND resolver can be imported wholesale into NetBSD at
any time (and by any programmer with a copy of the source tree).  Anyone
would then even be able to switch NetBSD to using IRS instead of
nsswitch with a simple recompile.

Imagine now nicely NetBSD's current NIH problem vanishes in a puff of
smoke when we can say that we can share the BIND resolver code wholesale
and even better that with a simple re-compile we can also support BIND's
IRS features directly in NetBSD!

-- 
								Greg A. Woods

+1 416 218-0098;            <g.a.woods@ieee.org>;           <woods@robohack.ca>
Planix, Inc. <woods@planix.com>; VE3TCP; Secrets of the Weird <woods@weird.com>