xml-dev - Re: XML Schemas: Best Practices

["Thomas B. Passin" <tpassin@home.com>]

Michael Brennan

> > From: Thomas B. Passin [mailto:tpassin@home.com]
> >
> > That's why I advocate that actual namespaces ought to be URIs that
are
> > clearly not URLs.  That would discourage people from trying to treat
> > them as if they were.  Especially in examples.
>
> I agree, but it doesn't seem to me that the prevailing standards give
> implementors very good choices in this regard. This whole
URI/URL/URN/XML
> Namespaces issue seems quite murky and confusing to me. It is
extremely
> counter-intuitive and confusing to the layman.
>
I tried out an xslt transformation using an arbitrary string for a
namespace - even without a scheme part - and it worked fine (I forget
which processor I used).  That's when I really appreciated that we are
essentially dealing with an arbitrary string (yes, URNs are supposed to
have a scheme part, but it's still arbitrary).
> ...
> This one single issue is such a huge source of confusion and
> misunderstanding that I really wish there was better guidance on this.
Just
> try to explain to someone who is not a regular xml-dev reader that
> "http://whatever..." is just a name and doesn't necessarily point to
> anything. All of the semantic arguments to this effect may be correct,
but
> they don't contribute to a good general understanding or making XML
> standards approachable to novices (whose understanding of the web is
shaped
> by their use of web browsers, and to whom a URL is an "Address" or
> "Location").
>
> Right now, our plan is to abandon our URN approach and move to URLs,
> following the lead of recent standards from the W3C and industry
> consortiums. If there is a better way, I would love to hear about it.
And I
> would add that I hate the various schemes I've seen proposed for URIs
based
> on UUIDs or similar schemes. Domain names are something that everyone
> (pretty much) understands at this point, and I want a naming mechanism
that
> builds upon DNS. Maybe we need a new "name:" URI scheme?

Yes, that would be good.

Personally, I don't think it matters at this point whether any urn
schemes and domains (if that is what they are) are "officially"
registered or not, for the purposes of namespaces.  I'd think about
making up a scheme name.  But if you want an "official" scheme, why not
use data: or gopher: ?  data: is perfectly legal, I think - there's an
rfc for it - and I bet no one is going to try to dereference a gopher:
url!  Maybe there are some other obscure schemes.

Plus, even a url can be "opaque", meaning that it is not supposed to be
parsed by the standard url processor beyond the scheme - according to
the rfc for uris.  So you could actually put anything into (at least
some) schemes.  I wouldn't worry for a second that I'd get a conflict
for

data:tbp15:ns:ontology:type:transformation:trans2531

I bet you could copyright a urn if you really wanted to (oh, boy,
that'll bring them out of the woodwork!).  That might give you some
control over others using it for unauthorized purposes.  Might do until
there's something better.

Cheers,

Tom P