RE: Enlightenment via avoiding the T-word

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

While munching on the grass, it is every cows' 
privilege to moo. :-)

As said before, it is the way of groves that 
makes them interesting.   It seems this 
saidToBeToComplicated approach is more elegant 
and robust than the DidLessSoCan'tDoEnough 
alternatives.

Still, I think it is semantic sharing that 
is the real bear.  Neither computers nor 
humans are dependable mind readers and the 
well-formed only approach requires that. 
I sympathize with what the well-formed only 
folks say, but it won't work outside a 
limited context of exchange or one in 
which all parties are under one authority. 
(Who chooses choices dominates design.)

So far, the most sensible bit I've seen 
in this thread has been Schemas are not 
well-layered with regards to type checking 
and name validation. I don't seem to 
get into trouble until I attach a namespace 
given that I do globalize the names 
and as Rick says, rename colliders not 
because it is theoretically elegant, 
but because it is easy to do and 
easier to document.   When does this fail?
When I don't get to control the names 
and have to exchange blind.  Then something 
has to give.  I can also localize in 
the schema and then I only get into trouble 
if they don't use my schema.  

Len 
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h


-----Original Message-----
From: Leigh Dodds [mailto:ldodds@ingenta.com]

I knew if I read down this thread far enough Len would 
mention groves! And if he didn't I was going to...

It seems like if you generalise Michael's notions of 
attaching generic annotations to the results of a 
parse you have a grove. And picking out the parts 
relevant to a particular application (or class of 
application) is a grove plan.

I know there is a lot of baggage attached to that 
particular term, but I've seen suggestion of a similar 
mechanism appear time and again. Even in Simpleton circles.