[
Lists Home |
Date Index |
Thread Index
]
On Wed, 27 Oct 2004 01:44:04 +0100, Bill de hÓra
<bill.dehora@propylon.com> wrote:
> Prediction: whatever replaces XML will look something like RNC or
> YAML - it won't be binary.
Agreed, at least tentatively.
I see the top problems with XML as:
- too big/slow for certain important niches (especially wireless)
- too difficult for ordinary mortals to correctly produce (cf Sam
Ruby's XMLDevCon paper)
- nasty impedance mismatch between DOM and XPath/XQuery data models
(this gets at the namespace problems others talked about)
- too much cruft was added for political compromises involving players
who are no longer around or have learned the error of their ways, but
lives on as "junk DNA".
Obviously this reflects my datamodel and DOM centrism, but I see the
way forward as accepting something very much like XML 1.0 as the
standard serialization that everyone claiming to be XML MUST support,
somehow aligning the implicit data models of the APIs on one hand and
XPath/XSLT/XQuery on the other, and gradually adopting a very small
number of alternative serializations as they prove useful in various
niches.
Something much like YAML or RNC is likely to be the ordinary-mortal
readable/writable serialization. Since both are close to
well-understood program language syntaxes, maybe it will also hit the
sweet spot between speed, readability, compactness. If not, maybe one
of the binary formats will fit the bill in the "speed | size is most
important" niches. Whether either a human readable or efficient
format is standardized is a question for the future when we know
more, not now.
Oh brother, I've drifted into the stepmother of permathreads, XML
datamodels and serializations. Sorry!
|
