[ On Sunday, January 28, 2001 at 10:23:45 ( +0900), itojun@iijlab.net wrote: ]
> Subject: Re: CVS commit: pkgsrc
>
> >> Removed Files:
> >> 	pkgsrc/net/bind8/patches: patch-ac
> >I think unless you've disabled the building of the deprecated 'host'
> >utility in some other way that this is wrong.
> 
> 	well,
> 	- i can't test conflict with nonexisting package

the original patch to disable 'host' was in the package because of the
conflict, even though it's been a "virtual" conflict for some time -- it
should have been left there...

> 	- no reason was given for removal of "host"
> 	so i simplified it with reason (less local patch = better).
> 	you are not on MAINTAINER so i did not bother to ask, sorry.

I can only supply so much information in a PR.  If I had commit access
I'd have already replaced the in-tree version of 'host' and then the
correct solution would be more obvious.

> >The 'host' utility must be a separate package made for the *real* 'host'
> >as the ancient BIND version would conflict.  I.e. the one at:
> >	ftp://ftp.nikhef.nl/pub/network/host.tar.Z
> 
> 	the behavior is not compatible, so i can imagine some prefers
> 	BIND8 host, and some prefer nikhef host.  please create BIND8
> 	host tool too.

It depends on what you mean by "not compatible".

W.r.t. command-line options (the things fingers form habits over), they
are in fact very compatible, and the behavior of each old option is
almost (with, I think, the exception of '-a') identical in the new
version.  The new 'host' is after all a direct decendent of the ancient
one, and meant to be upwards compatible.

The ancient version of host is pretty much useless for anything but
looking up plain records, so of course the new one is going to be
"incompatible" in many ways since it offers so many new features and
options.

I'd be extremely surprised if anyone's scripts depended on the exact
format spit out by the ancient version, except the 'mxlookup' and
'nslookup' scripts included with it, but since they're also updated this
is a moot point.  The important information is still the last field, so
that much hasn't changed.

As for human preferences, well I can't even imagine any complaints about
output format for humans as being valid in any way -- humans can adapt
in the blink of an eye, and if they don't then they know how to read the
manual page to get what they want.  Is there really any significant
difference for a human?  Who cares?  Has anyone ever said they prefer
the output of the ancient one over the new one, or are you just
imagining that they might?

I'm not sure what you mean by a "create a BIND8 host tool too".  If you
mean you'd like to see a "pkgsrc/net/host-ancient" package, then there's
no point -- it's ancient and already deprecated in future versions of
BIND (despite the fact that some fools find the time to hack on it and
patches occasionally make it back into updates of previous versions).
I've no time to support software I haven't installed or used for over
half a decade (and it's *much* older than that, 1983)!

-- 
							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>