RE: Internationalization Requirements

I really think one has to keep in mind that transparent content negotiation
goes way beyond internationalization.  The spec is a little scary in the sense
that a client and a server can collude to use *any* attribute, even one that no
one else in the world knows or cares about, to choose a variant.  I also think
you have to keep in mind that there are different relationships possible
between the backend variants gathered under the same URL.

It may be that there is one master variant and all the others are automatically
derived from it.  For example, the prime variant may be SGML and from it flow
HTML and PDF versions.  Checking out, editing, and checking in the secondary
variants in this model is senseless.

It may be that there are several co-equal variants that must be updated
independently and manually.  For example, each variant is in a different
language and one doesn't trust automatic translation software.  

It may be that there is some sort of hybrid: different content type variants
for each language variant, all under the same master URL.

I've come to the conclusion that as long as it is possible to get at the
individual variants when necessary, and determine exactly what variants there
are, and to not check out an URL that has variants under it (because one could
end up in the bizzare position of using one set of client settings to check it
out and another to check it in and having the wrong variant get updated),
it is really up to the application built on top of the protocol to sort
this all out. 

	-- Mary
	   Holstege@firstfloor.com

| Mary Holstege, PhD
| FirstFloor Software, Inc.               (415) 254-5161
| 444 Castro Street, Suite 200            (415) 968-1193 (Fax)
| Mountain View, CA 94041                 http://www.firstfloor.com/

Received on Friday, 25 July 1997 11:19:11 UTC