[ On Sunday, January 28, 2001 at 14:37:19 ( +0900), itojun@iijlab.net wrote: ]
> Subject: Re: CVS commit: pkgsrc
>
> 	i'm not in love with the tool (i always use dig), i'm just saying
> 	i don't follow your argument.  for me not shipping certain
> 	tool in a package needs a rock solid reason, your argument does not
> 	really sound to me a reason for keeping local patch around (pkgsrc
> 	maintenance cost is biting us so I'll do virtually anything to decrease
> 	the number of costmetic or reason-less diffs).

There real maintenance headache is with keeping an ancient version of
'host' in the BIND tree!  ;-)

Also note that BIND doesn't *need* "host" either (nor dig nor nslookup
for that matter, though having at least *one* of them is handy).

Traditionally the ancient version has always been in the 'tools'
sub-directory and the new version has been in the "contrib"
sub-directory.  Unfortunately with the advent of BIND-8 the ancient
version was "promoted" to the 'bin' directory and the new version was
left out.  My understanding of the decision that lead to this was that
since the new version was being independently maintained it wasn't seen
as a good idea to include a stale copy of it in the BIND distribution.
I'm not sure exactly why it was felt the ancient version should be kept
instead, since it's most stale of all.

'host', BTW, was initially written as a "unix-like" tool to replace the
more interactive 'nslookup' (and indeed includes some code from there).

> 	next time, you may want to put one-line comment at the top of patch-ac,
> 	following $NetBSD$ tag at the top.

Hmm...  OK, though I'm not so sure I've seen any documented guidelines
suggesting this practice....

> 	i can't seem to make them behave equally.  are we seeing the same tool?
> 
> (BIND8 host)
> % /usr/bin/host www.netbsd.org
> www.netbsd.org is a nickname for nbwww.isc.org
> nbwww.isc.org has address 204.152.186.171
> 
> (new-host)
> % /tmp/h/host www.netbsd.org
> www.netbsd.org          CNAME   nbwww.isc.org
> nbwww.isc.org           A       204.152.186.171

And what's so different about that?  They've both behaved exactly the
same way and they both told you exactly the same thing using exactly the
same command-line!  The new version has just been more concise in its
output format, but anyone not understanding the new output format won't
know what they're reading from the ancient version anyway.

> 	well, you are the only one I've seen who is vocal about this particular
> 	preference about new-host and BIND8-host.

I've made noises in the past in the bind-workers list, but usually I
simply try to promote the new version of 'host' instead of actively
campaining for the replacement or erradication of the ancient version in
the BIND distribution.

My preference is only for the new and indispensable features.  Any
differences in the output format are irrelevant to me so long as the
necessary information is presented in some reasonable way.

I couldn't function from day to day without the '-A' or '-C' options,
and I often use '-Z' to see the zone-file format records, especially if
I might have to cut&paste for fixes or evidence.  The sanity checks in
the new version are far superior too.

I see that the '-C' option included in the rewrite of the ancient
version in the BIND-9 distribution, but in an incompatible fashion
("host -C ."  doesn't work, for example, and it does not do all of the
necessary checks properly, so it's pretty much useless since it needs to
be wrapped by a script anyway and a script could do the proper check
with only the most basic RR retrieval tool).

There's even some support in the new version of 'host' for IPv6 too (and
it seems some minor support's been put into the ancient version rewrite
for the BIND-9 releases too).

I know directly from many other people that I'm not the only person who
greatly prefers the new version of 'host'.  I'm probably more vocal
about it than any of them though....

>  all the latest releases of
> 	*BSD ship with either BIND4 or BIND8 host in /usr/bin/host, i have
> 	never heard of plans to replace them.

That's too bad -- they're missing out on making some very powerful and
incredibly useful diagnostic tools to their users.  I've been installing
and shipping the new 'host' on systems I build for quite a few years now
(6?), and those of my users who do DNS work every day are very happy
about that.

> 	which future version of BIND are you talking about?  could you clarify?

I don't seem to have saved the announcement, but IIRC someone from ISC
has stated that 'dig' will supersede both nslookup and the ancient
version of 'host' as the "official" DNS test and query tool for the BIND
distribution.

I personally can't imagine why 'dig' was chosen over the new version of
'host' though since 'dig' is about the most useless and non-unix-like
tool I've ever seen, other than maybe nslookup, though even nslookup has
a usable command-line.  I've never used it and I personally delete it
from all my own systems and don't normally build it in my source trees.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>