Re: [xml-dev] Here's a good question

[Rick Jelliffe <ricko@allette.com.au>]

 From: "Don Park" <donpark@docuverse.com>
 
> Ultimately, it was the DOM requirements that neutered it.

Yes, as usual Don raises a worthwhile point.

XML Schemas is another example where innocent requirements
sets the big ship on a course that cannot be changed easily.

By contrast, XML's 10 "goals" had enough motherhood statements
(and it was a trimming job) to trim out a lot of fat: in particular, 
"terseness is of minimal importance"

I would say that for requirements to have good fruit, they need
to avoid statements which prevent moderation: this is not just
the extremes of anorexia or bulemia, but extremes of 
abstraction and too-thin or too-thick layers.  

I think there are two things that might be very useful for standards-writers:

 1) The "worse is better" approach that we will put out a minimal spec
   first, then a compatible more complete one second: the SAX 1
   being followed by SAX 2, for example. So, for example, 
   XML Schemas datatypes should be two specs: one for primitive datatypes
   and one later one for type restriction mechanisms.
  
 2) The first minimal spec must be fully usable by the main use of the expected
    layer that sits on top of it.  So, for example, XPointer should not have non-WF 
   ranges because  these are not simply useable by, in particular, XLinks used for
   retrieving data over the web.  

 

Cheers
Rick Jelliffe