RE: "industry needs"

The point here was not to debate the details and perceptions of of MathML
though I think it and the process that was followed does provide good
counter examples to some of the ideas I have been hearing. 

Instead I'd like us to focus on

1) The importance of rules as data, and facilitating their archiving and
exchange.
2) NOT accidently or deliberately ruling out a major class of rules.
3) Providing an appropriate enabling infrastructure

Recommendations are very hard to come by.  There is a heavy responsibility
in developing a 
recommendation to make it count as you can't keep going back and changing
them, or adding new ones (at least with any concsience.)  

Just because we don't use rule class X or Y does not mean they should not be
exchanged between those who understand one class or the other.  Just because
we don't know how to interact with rule class X or Y does not mean no-one
should be able to write such expressions or exchange them.

A key point (and lesson from MathML) is just because you allow someone to
write something does not mean that you require everyone to implement it.
They need to be able to recognize it for what it is and abort or hand it
off. It does become usable for those who can. "Not trying to evaluate
things" is an important part of that exchange.

It should be an enabling technology in the sense that it doesn't prevent
those who have figured out how to mix the classes of rules (including nice
formal consistent models yet to be developed) from actually doing so.

The whole issue of re-writing procedural rules as declarative is, I believe,
outside this framework, though such activity would/should benefit
considerably from a framework that could write both.  It would be overly
burdensome on the procedural rule family.  (Though I would love to see it
formalized. )



Stan

PS.  Of course we might disagree on the facts.  :)

Received on Tuesday, 6 June 2006 15:29:26 UTC