RE: Defined sets, accept sets, and <banana> elements

Dave Orchard writes:

> Tremendous!  This is wonderful.

Well, thank you!

> Do you think this would be worthwhile to add to the versioning 
> works, or is it extra material that takes away from the larger picture?

Actually, I think it suggests something a bit different.  The short 
summary of what I concluded is:  defined-set/accept-set adds a lot of 
value in the case where V1 says that extension content is truly ignored; 
as best I can tell it doesn't help much in the case where V1 applies some 
other default semantic to extension content.  So, if I were setting 
priorities, I would say that the next step would be to find out how 
important such languages are to users.

My intuition is that it's very, very common for languages to provide a 
semantic other than "just ignore";  while HTML itself does use the ignore 
rule, neither HTML+DOM nor HTML+CSS do.  I strongly suspect that many 
vertical languages also make extension content available to applications 
(much as the DOM exposes <banana> elements for HTML), and so on.  Insofar 
as I'm right about that, I think we want to ask ourselves whether we're 
happy with a core model that seems to say very little about these sorts of 
extensible languages.  If we're still OK with defined/accept, then we 
should ask whether we also want to tell a story about HTML+CSS, and then 
finally we can get to the question you're asking, which is whether 
something derived from the content of my email belongs as an addition to 
the finding as it stands.  I'd certainly be glad to see it included if 
it's helpful.

FWIW:  I still believe that there are alternatives to 
defined-set/accept-set that would handle in a first class way the many 
languages in which open content does have a nontrivial semantic.  If I 
were writing the finding just for my own satisfaction, I would be tempted 
to explore those models first, and see whether the "ignore extensions" 
languages don't just fall out as a special case.  Still, if the rest of 
the TAG is happy with the compromises inherent in defined-set/accept-set, 
I don't think I'd stand in the way.  It's a nice, clean model for the 
"ignore" languages.  Still, I'm not convinced that define-set/accept-set 
sheds very much light on the cases that I think tend to be most 
misunderstood by users.  So, let's see what everyone else thinks.

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------

Received on Wednesday, 20 June 2007 21:20:57 UTC