[ On Wed, November 18, 1998 at 14:54:52 (-0800), Bill Studenmund wrote: ]
> Subject: Re: mysql: hostname 
>
> > Of course you can use the domainname for both on systems that use both,
> > provided that you're also sane enough [;-)] to use the same string for
> > both purposes....
> 
> What if you can't? I have administered one cluster, and will soon be
> helping set up another, where domainname != DNS domain name. Here at
> Stanford, all the computers on campus are in the stanford.edu DNS domain.
> So we can't set the two the same. :-) (well, one YP cluster per subnet
> could, but then the others on the same wire couldn't).

Well, yeah, I suppose people do get stuck with legacy systems, and yes,
I do realize there's enough utility in the NIS interpretation of a
"domain" where it might not make sense to have as many DNS domains
(though it shouldn't be impossible, especially since DNS domains are
quite flexible and can overlap).

> What about hosts which use both?
> 
> I guess part of this question stems from the fact that I never understood
> why folks would want to get host name resolution via YP.

Me neither, which is why I don't have an answer to your question.

> And that's not a
> question I expect anyone here to answer. :-) I mean DNS is great for host
> name resolution, and yp is fine (*) for keeping user ID's consistent
> across hosts, so it seems to me it'd be best to use each for what it's
> best at.

But even in this situation it's not too difficult to imagine a scenario
where the NIS "domain" name turns out to be unique and thus different
from the DNS "domain" name.

> But regardless of anyone's thoughts on what I describe above, I really
> think we sould not recomend using "domainname" as the DNS domain name. I'm
> afraid new folks could get tripped up. "domainname" should always be the
> YP domain name (which can be equal to the DNS domain name due to a policy
> decision, but it's still the YP domain name).

I'll agree so long as "we're" promoting NIS/YP.  In general I wouldn't
want to do that -- it's necessary in the short term where it's already
in use, but I wouldn't start anew with YP.

I've been on a long, sometimes silent, sometimes not-so-silent, campaign
to "subvert" the use of the kernel domainname for the purposes of
systems that reside entirely within the DNS (no YP in sight!).  I think
there's just enough of a significant advantage in doing this to make it
worthwhile (esp. with mobile computing on the horizon).  However in the
end it may make more sense to have a kernel "dnsdomain" value that can
be used to centralize this value, and leave "domainname" for the NIS/YP
folks.

-- 
							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>