[
Lists Home |
Date Index |
Thread Index
]
From: "Dare Obasanjo" <dareo@microsoft.com>
> S-exprs are equivalent to XML for a large class of XML users.
So what? XML has succeeded for a bundle of reasons, while S-expressions have comprehensively failed (and I speak as someone
who co-wrote a LISP, worked supporting TI Explorers which had LISP as its system language, and love eval and S-expressions.) Who can rule out that the things in XML not found in S-expressions helped it reach a critical mass?
> First of all if you think about why most people use XML you'll
> realize it's mostly because of the network effect and not anything
> intrinsically fantastic about angle brackets.
So the essense of XML is the herd instinct? ;-) In any case, after the
technology has been implemented by developers, other people are
using products not XML: so what the users of XML-hiding products
think about XML is hardly likely to reveal much about XML.
For angle brackets, I think the evidence is mildly against you. In SGML people can change delimiters: almost none do (unlike "&" which is sometimes remapped) --the need to remap delimiters had become moot by the time XML discussions came along.
There *is* something quite graphically attractive about angle brackets used for tag delimiters, it seems: first, they are not used in general text so their status is visually clearer (contrast () and to a lesser extent {} or []), second they visually group the tag (especially small tags). There may
be even better graphical characters for tag open and close, but not in ASCII, to my eye. (This is something that would better be addressed
by ethnography and psychological experiments, rather than pure reason.)
> From that perspective, most users of XML could just be as well served with S-expressions which for all intents and purposes would be equivalent to XML for their use cases.
Err, if XML helps users regard encoding as a peripheral issue, then
that is a sign that XML looks after it quite well, not that it is inessential.
Saying that the encoding is an integral part of the data format that should
be labelled is not something obvious. The RELAX NG compact syntax,
for example, does not provide any way to label encoding, and I see Jing has command-line arguments to accept different encodings, which does not scale well.
Cheers
Rick Jelliffe
|
