OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help



   RE: [xml-dev] Re: XSLT and the 80/20 point (was RE: Tim Bray on " Which

[ Lists Home | Date Index | Thread Index ]

Yes, and in their day and on the system for which they 
are re-architected, they thrive and dominate.  That's 
directed evolution in action.

Yet hindsight has to be as accurate as we can make it, 
or we make the wrong decisions.  Again, there are conceptual 
and systemic issues for the technical decisions.  SGML is 
big by comparison to XML.  The re-architecting to make 
SGML work on the web was based on decisions that are themselves 
being reviewed.  XML fixes an SGML Declaration in code.  Wise? 
For a very large percentage of XML developers, yes.  For some, 
no.  Now the question becomes does XML get larger to meet the 
needs of those who have a use for the declaration, or do we 
say, "no, that's SGML and you should use the right tool". 
Then the responsibility to upgrade SGML in response to that 
user's needs is clear and XML stays targeted.

But if we go down a path of "SGML doesn't matter; SGML is 
document-centric and we all KNOW that data centric is the 
way", then XML becomes bastardized.   SGML like any other 
standard targeted to too broad a set of surface 
similar applications becomes too big.  Do that to XML and 
we miss the value of the lessons of history, but how 
to stay simple and stay pertinent is hard.  We could restrict 
the content scope (somewhat like saying the WWW should be 
English-only) and get a simplification.  Anyone really 
want that?  Some did but it's stupid on the face of it. 
It murders the reason for having the web to begin with: 
to create, preserve, and disseminate knowledge.

So where we simplify matters.  To make that decision, 
we need a clean history and it must include both conceptual 
and systemic information.  It's no good to say "SGML 
doesn't matter" if we don't put that in a systemic 
context.  XML's design would not have mattered in 
1986.  It would have been understood as too loose. 
Validation mattered then.  Character set descriptions 
at the document locale mattered.  Memory storage size 
mattered.  Depth of recursion mattered.  EBCDIC mattered. 

Hypertext didn't matter.  The Internet didn't matter.

Five years later, when hypertext became something more 
than a lab curiosity and desktops became more than 
toys, it was time to rethink that.  Five years after 
that with some experience with a very large distributed 
hypertext, the lessons were clear.  At that point, 
many of the SGML Way decisions clearly did not apply 
any longer for THAT SYSTEM in the majority of uses. 
Now almost five years after that, some more rethinking 
is happening.   As a result of that, we may see more 
SGML-like features reemerging.   Should they be in 
XML or should SGML be reworked with a seamless path 
between the two factored in as has been done to date?

In hindsight, what are the lessons, what are the 
requirements, and what are the ways forward?  The 
human dimension of this is that we have to be extremely 
cognizant, and I for one do not trust those decisions 
to be made by any person or group who thinks that revisionism 
of history is needed to sell their case.   If we go 
down that path, we simply repeat mistakes.  Clarity matters.


From: Jonathan Robie [mailto:jonathan.robie@softwareag.com]

At 01:06 PM 3/18/2002 -0600, Bullard, Claude L (Len) wrote:
>Except in hindsight.  Everyone is an expert at that.
>XML was designed "in hindsight".
>Hindsight matters.

There's a lot of truth to that.

Everyone knows about Occam's Razor, and most designers say simplicity, 
conceptual integrity, etc. are important to them. Still, most of our 
initial designs are more complex than we would like. If we're lucky, our 
colleagues point out potential simplifications. More likely, they point out 
all the important cases that our initial solution did not cover. A good 
designer then goes back to the drawing board, looking for a solution 
general enough to cover these cases without a bunch of special casing, 
using hindsight as a tool for knowing what to throw out (grin!).

One set of technologies that matter are the break-through technologies that 
show how something new can be done. These generally have warts, but they 
show the way. Ten years later, we may want to use something else. Another 
set of technologies that matter are less brilliant, simple reworkings of 
the original ideas. XML was not creative or original. Neither was Java, 
really. Both are excellent examples of hindsight in action.


News | XML in Industry | Calendar | XML Registry
Marketplace | Resources | MyXML.org | Sponsors | Privacy Statement

Copyright 2001 XML.org. This site is hosted by OASIS