Re: Proposal: requesting XML records

Alan Kent wrote:

>     "If Comp-spec is used, this agreement does not apply."
>
> Is the purpose to say Comp-spec is out of scope (not defined) or that
> it is recommended that Comp-spec *should not* support these set names.

"Out-of-scope".  The proposed agreement simply would not apply to compSpec
and there is no intent to suggest that compSpec should or should not support
this.


> Another observation (not a problem, just an observation), from memory
> element set names are not case sensitive (in a quick skim of standard
> I could not find this just now, but I recall seeing it previously).
> URIs on the other hand are case sensitive. I guess there is no issue
> as long as clients are told to always supply the URI with the correct
> case. (If they don't, should it still work - maybe just undefined).
> But as I said, probably not an issue in practice so not worth mentioning.

Yes, ESNs are case insensitive and URIs are case sensitive.

So, suppose "http://www. ........... xyz"  and "http://www. ........... XYZ"
identify two different schemas. Suppose a server knows the first, and the
client sends the second. The server will treat it as the first. The only
reasonable answer to this problem is that this never should have happened to
begin with (these two names identifying different things). But how do you
prevent that from happening?

I think the practical answer is this:  I think that the domain-name part of
an http uri is case insensitive, at least in practice. For example,
HTTP://WWW.LOC.GOV resolves to http://www.loc.gov, whereas
HTTP://WWW.LOC.GOV/Z3950 does not  resolves to http://www.loc.gov/z3950.

So, assuming that
(1) the naming authority part of a uri is, in practice, case insensitive; and

(2) a single naming authority has enough sense not to assign conflicting
names;
then I think the problem should not arise.

--Ray

Received on Wednesday, 2 July 2003 15:58:33 UTC