[Ex-plain] XML Schema for 1.3
mike at miketaylor.org.uk
Fri Mar 15 21:39:21 CET 2002
> Date: Fri, 15 Mar 2002 15:20:21 -0000
> From: "Matthew Dovey" <matthew.dovey at las.ox.ac.uk>
> > So if we make host required and not repeatable, is everyone happy
> > with allowing either numeric or symbolic for its contents?
> > Otherwise it has to be repeatable, I think.
> I would support this.
> As for the issue about search using the host attribute (is that
> attribute 1?) - I think it is up to the individual implementation of the
> explain-- Z39.50 server to determine how to cope. It could simply search
> what is in the explain records it has been given (harvested etc.) It
> could attempt on the fly DNS resolution. It could cache DNS resolutions
> for performance. It could just check for alternative DNS entries when
> the record is added to the database but never again etc. Anyone ready
> for an implementors agreement ;-)
I know this was a joke (yes, I was able to read the ";-)") but I think
Ralph's recent comment bears repeating: to paraphrase him, ever IA is
a tacit admission that we badgered up in the original definition.
On the single/multiple hostnames issue, my feeling is that I have no
great love for multiple hostnames, but if Rob really wants them, then
I don't feel motivated to stand in his way. I take Ashley's point
that for his servers, listing both canonical and preferred hostnames
would be actively confusing -- but of course nothing obliges him to do
so. Even if we stick with the current, multiple-hostnames-are-OK,
version of the DTD, that doesn't _oblige_ anyone to list multiple
hostnames (and I know I never will.)
If we do decide one hostname is the correct answer, then I oppose this
idea that we should accept either IPs or names. Name are right.
Names are good. IP addresses are for system administrators, and no
mortal being has any legitimate need or right to mess with them. I
have zero sympathy with Rob's globally-accessible but unnamed server:
he should just give it a name and have done. And as Matthew points
out (in an earlier message), users whose DNS servers have gone away
have more pressing concerns that following F&N links!
Summary: I prefer One Hostname, symbolic only; but I will not fight a
consensus that multiple hostnames are OK. I would like to see a note
in the DTD's commentary to the effect that the primary hostname should
be a symbolic one where possible.
/o ) \/ Mike Taylor <mike at miketaylor.org.uk> www.miketaylor.org.uk
)_v__/\ "Even the genus _Tyrannosaurus_ hasn't yet been given a
good diagnosis. Right now, its diagnosis seems to be ``big
tyrannosaurid from North America''" -- George Olshevsky
More information about the Ex-plain