Lists Home |
Date Index |
Right and ok by me. But it doesn't have a high reliability
metric when one has to choose which of the local interpretions
is right, testable, buyable, get's us out of the Factory
Acceptance Test and into the field sort of thing. Even
where there are data standards, local implementation is
good for local implementors. Selling a shrinkwrap to a
set of local requirements keeps the costs high. Remember,
we had very advanced hypermedia applications before HTML
and Mosaic. The IETM business couldn't take off because
no local buyer interpreted the requirements the same way.
Even TODAY, they are still trying to come up with a
common player platform for all services, and this, some six
years after the work on the MID stopped for all intents
and purposes. They still don't buy a web system as an
IETM common platform.
The problem, I think in retrospect, with
the ideas of Andreesen et al was that they a) didn't
know they weren't doing something advanced b) believed
that theirs was the only system that would count. As
soon as they had competitors, they had trouble.
From: Simon St.Laurent [mailto:firstname.lastname@example.org]
Len Bullard writes:
> One can always make a tag soup work in a closed application.
> The problem is across applications, the kinds of things that
> arch forms were originally proposed for. Note that I am
> saying "for any given problem in every case". In other
> words, the concept of complete self-describing types
> begins to break down if these types have to integrate
> with more than one application. IOW, the concepts that
> lead to the namespaces solutions only cover a subset of the
> problems they are proposed for. Then they begin to
> break down.
It depends on how you interpret "work". If you treat document
interpretation as based on local understandings, not some global vision
of agreement, then tag soup is fine. Local processors will figure out
what they can, just as they would have to do so anyway.
Namespaces provide extra information to that process, and so do
architectural forms, and so do schemas of whatever type. Depends on
what you want.
Given your usual arguments, I doubt that's what you'd understand by
"work", but it does just fine for some of us.