RE: RDFS versioning

: Aaron Swartz:
:Ahh yes -- if that were the only issue, I wouldn't be upset. 
:But the spec
:also "recommends" that RDF-based systems can cache a schema 
:indefinitely
:(which I read as "until the next Ice Age") which means that you're
:effectively stuck:

Well that's fine and consistent with the change schema == change URI.
Ignoring typos and discovery issues for now, imagine A and B are machines
chugging along happily using schema a, which both have cached. The authors
of schema a want to update it, but to do this and be good web citizens, they
have to give it a new URI. Let's call that new version schema aa. 

Since in this case, the semantics of schema a aren't changed without
changing the URI neither A or B can get confused, when either of them checks
a for a new version. In fact they *never* need to check because the authors
by being good citizens are effectively stating that any schema they publish
will never be changed. So there is no caching anomaly to worry about.  

Looked at in a certain way, this is a very, very good thing. If A moves to
schema aa and expects to be able to converse with B using schema a, that's
not the fault of the publishers: for all intents and purposes A is trying to
communicate with B across two different schemas.

That's why it's far from a silly requirement. If anything the wg should
really consider upgrading change schema == change URI to a MUST. 

Bill de hOra

Received on Thursday, 22 February 2001 10:14:51 UTC