Which Choices Matter? (WAS RE: SV: [xml-dev] Tim Bray on "Which Technologies Matter?")

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

Moving this to a different plane, I say the choice of means of choosing 
matters.  What happens if there is only one choice?

Consider that the momentum choice (the means of choosing) has 
chosen Internet Explorer.  Ok.  What about the choice of application 
language?  IE isn't a conforming XML application language 
system, is it?  Well, can we tell quickly?   Sure.  Give it a 
table with all of the </tr>s and only one <tr> in the first 
row.  Keep all the rows balanced for <td></td>.  What does 
it do?  It looks perfect.  What happened?  IAW the choice 
"to be liberal in what it accepts", it fixed it for you.   

What is the problem here?  

The Draconian Parse:  XML's one inviolate rule for 
XML processors.   BTW, it took <a name= > without a gripe 
too.  The only reason the person who found this did is because 
it didn't like a basefont statement that was well-formed 
but unimplemented.  Even then, it only showed up because 
she tried the Print Preview and serendipity got the hindmost.

OTW, it was on its way to the loading dock.

Say, OK, but given a DTD and a validator I can find that. 
Sure, but do you have one and did you?  What if a generator 
is being used to produce that and the only quality step 
was inspection in the rendering media engine? Some 
designers really trust their code and discount the rest of the 
XML systems above the level of syntax.

Different choices matter.  That is the problem of the 
80/20 decision.  Without hindsight, it is rather easy 
to make a bad choice of means and watch the other 
choices cascade until serendipity or disaster strike.

XML sits squarely on top of SGML as a subset, and HTML 
and SGML for hindsight.   Good choices, I'd say.

len