Abstraction vs Scaling (Was: Native XML Interfaces)

[cbullard@hiwaay.net]

Lauren writes: "although HyTime to XLink didn't quite pan out."

Dan Connoly wrote some years ago that the problem of SGML on the Web  
was it wouldn't scale.  Some interesting evidence lately encountered.

1.  XLink is still widely used in IETM standards.
2.  HyTime archforms are still referenced and implemented in some IETM  
applications.

XLink may not have fared well on the web but it was drawn into the  
applications where its parent was adopted and still survives like a  
coelacanth in the deep dark waters where only the toughest creatures  
live.  Ironic?

Scale matters to the web.  In other applications it is not scale but  
precision of naming and identitying (because no, they aren't the same  
thing) matters.  This is where Kurt has a point:  relating information  
as a graph and organizing it into a tree that makes it easier for the  
human to consume are distinctly different problems.   Organizing an  
application to scale and making it fast and precise are distinctly  
different problems.  REST is a bridge but that and only that. Beneath  
the abstractions that make the standard referentially tight is simply  
a network emailing state.

At one point in its evolution, design, whatever, it was in the  
Goldilocks zone:  about the right set of concepts for the best  
scaling.  Then it became more abstract, it baked too long before being  
implemented widely enough to figure out what the next version should  
be and how much and what should be taken away.  DOM thrived longer  
because at the time of the launch there was a social.technical need  
with lots of customers.  Timing is everything in standards.

Lesson learned:  abstraction and scaling are not necessarily allies.   
What a programmer wants and what an author can do aren't allies unless  
they are required by treaty to be so.  This is why the slow stumbling  
work of standards committees go on.  Peace and prosperity.

len