[
Lists Home |
Date Index |
Thread Index
]
> From: Martin Klang [mailto:martin@hack.org]
> > See my posts to Dave. Starting of with a registry based on existing
> > short-string registries, along with subscoping (I'm
> changing my terminology
> > already!) it think addresses this issue.
>
> hey, if we allow for subscoping with slashes as well as dots,
> and add a
> protocol specifier, (okay so we're not really talking XML 1.0
> here) you
> could have elements with prefixes like:
> <http://www.xml.org/very/cool:element>
> funky eh! who needs namespaces now! (excuses to xml.org)
Yeah, yeah, yeah. Verrry funny.
> i can sort of see why you'd want to register prefixes
> somewhere ('cos yer
> lazy! (oops, sorry)), but imho you couldn't really do it without
> duplicating what the URI is - a unique identifier, right?
Yes, which is why I was very tentative about the proposal for subscoping.
Iff there's not enough short strings for namespace prefixes (which are ids),
and subscoping starts to make sense, then the arguments that we're back
where we started from make a lot of sense. The main intent for prefixes is
to preserve readability; subscoping would diminish that.
Are there enough short strings for everybody? Hmmm....
> on an ever so slightly more serious note,
Why start now?
> it seems to my
> humble self that
> a mechanism (i almost wrote 'standard') for namespace
> registries (for the
> URI part, not the prefix!) could be very useful indeed. say
> to locate the
> schema or wsdl that defines a document you're trying to read,
> or service
> you're trying to talk to.
> at the moment you can try to dereference the URI in the hope
> that it's not
> just a namespace but also a URL to the resource, but rarely
> will that give
> you anything but a 404. and it shouldn't really, anyhow.
> but the server part of a namespace could be used to send a request to
> resolve or get a list of resources.
I don't know... it seems to me if you want that sort of functionality, just
have the URI address an RDDL doc. No need for a registry. Maybe I miss your
point.
[snip]
> sorry for the lack of direction here, my point is simply -
> registry good, centralisation not necessary
Yeah. My point: centralization of authority good, centralization of
registry resources not necessary. At least from a functional aspect; from a
maintenance aspect it may well make sense to centralize the registry. If
it's static, you shouldn't need whole-scale replication; I think piecemeal,
local, as needed replication would be sufficient. For the most part you
could assume that a registered prefix (as determined by appearance) would in
fact be so. That's really all you would need to know for raw document
processing; meta resources would be something that could be supported
elsewhere.
You're first point, though, is very valid:
1) are there too few short strings to serve as registered prefixes now
served by existing (and future) URI-based namespace ids?
2) iff the answer to 1 is 'yes', then the second question is whether
subscoping rectifies the issue.
3) iff the answer to 2 is 'yes' (and it's the only reasonable approach),
then one can ask if subscoped prefixes are readable enough
4) if the answer to 3 is 'no', then what mechanism could be used to make
them readable?
5) if the answer to 4 is 'something along the lines of the namespace prefix
mechanism we have now', then the final question is 'why isn't this a stupid
proposal'?
That's the one series of arguments that would lead me to that there's no
hope of anything better than what we have now (aside from dumping default
namespaces altogether).
>
> /m
>
> Martin Klang
> http://www.o-xml.org - the object-oriented XML programming language
>
|
