[
Lists Home |
Date Index |
Thread Index
]
Saying it is "just politics" is maybe belittling.
It is need and that is always a local politic.
The problem is a local politic going global at
the point of the standards sword.
We said we wanted standards, told everyone
that was the only way the web could work,
embraced that the line best worn would be
the W3C line, and made household names of
the tailors, but come time to cut on the dotted line,
we don't all really want to wear the same
ones. Normal is as normal does, yes? What did
Megginson say:
"...what matters is what people actually do, not
what spec writers invent or marketing departments
announce support for." Of course, someone
claiming to do SAX that renames all of the
API calls might get a different answer.
Back when we were first proposing SGML to the
PDES crowd, at the end of a long day, they
concluded that the SGML type would be TEXT
and that was it. Just a text string.
Most of the PDES community considered that
too weak for a real document type and said so.
It turns out several years and many changes
of sides later, TEXT is as meta as meta gets.
So this discussion, or class warfare, has
gone on for a long time in markup. It comes
down to truly what is the least amount
of agreement required to get the most
work done without consultation. Syntax
and bits on the wire seem to be it. After
that, it is all special case pleading.
I don't object to datatypes. It is useful
to know what the intent (as in RULE) of the
sender is because I don't want to explore
a space by extension (enumeration).
That doesn't mean I have to
follow the RULEs, but I might. At the end of
the day, shy of infoset applications,
XML is just TEXT but woe unto the one
who does not read well the BNF or
tenderly care for the codepoints.
len
From: Jeff Lowery [mailto:jlowery@scenicsoft.com]
So let's see if I understand this:
Are XML datatypes core technology? No.
Are XML datatypes useful under certain contexts? Certainly.
Can the support and transport of datatype information impose a burden in
contexts where datatypes are not useful? Of course.
Therefore:
Datatype support should be modular.
Datatypes should not creep into core Recs.
In contexts where datatypes are not useful, there should be a mechanism to
toss away or filter out datatype noise. (I suppose the same could be said
for namespaces, as well, or any other meta-meta-tag info.)
So why is this class warfare and not merely an argument for good software
design?
Politics? Mmmm... could be.
|
