[
Lists Home |
Date Index |
Thread Index
]
It helps if those doing the simplification are roughly
the same people who did the original. In the case of
XML, that was 90% true.
One could argue that simplification of complex systems
is normally achieved by encapsulating a feature into
a different component or layer and decoupling it
via an interface. DTDs were made optional and that
made sense for networked systems. But the problems
of augmentation and the requirements for validation never
disappeared. The syntax problem (DTDs in a different
syntax) was factoring because in practice, it wasn't
a problem. It simplified the parse. Was that a large
gain? Opinions vary but it did open up the technology
to competing solutions there, and one might say it
simplified the technology but made the environment
more complex (how many validation types does one want
to support).
Namespaces seem to complexify everything they touch.
Because SGML did not deal with these except in subdocuments
where the complexification was rough encapsulation, there
was no prior experience except by analogy to other
languages and the Oster system.
If one accepts the assertion that XML itself
is only syntax, then namespaces work reasonably well.
If one attempts to use them with any other semantic,
results vary.
You have no doubt noted Sandro Hawke's thread on the TAG
concerning modularity and composable languages. Permathread
#3, (How Does XML Support Composable Languages)
is on the schedule again.
I wonder who will be the first to resurrect Brad Cox and
Software ICs. Now that is a golden oldie, but closer
to the solution than most. It leads to C#, Java and UML
and away from declarative data objects.
len
From: Simon St.Laurent [mailto:simonstl@simonstl.com]
clbullar@ingr.com (Bullard, Claude L (Len)) writes:
>At least it is easy to see now how SGML became as
>complex as it did over time. What doesn't get
>put into XML shows up in the applications. It
>will be fun as some of us get closer to retirement
>and watch the next generation of "we can do this
>simpler and better".
>
>Cars and jets never get simpler overall. Why
>do people believe software should?
In general, I don't expect it should. In this case, I think it's funny,
because a group threw away all those features, claiming they were too
obscure, and then reinvented them in a form that seems even more
obscure.
Reinventing the wheel is fine, as long as you come up with a better
wheel. I think XML 1.0 (and perhaps XSLT 1.0) set a standard for
improving wheels by simplification that hasn't been matched since.
|
