Lists Home |
Date Index |
1. Those slides are a minimal skeleton designed to hold up an
hour or so of verbiage. Having said that, they're in a public
place with my name on them so I guess I have to say whether I stand
behind them or not.
2. I stand behind them.
3. I guess someone should expand them into a real essay with flesh
between the bones.
4. The audience is a mainstream tech-conference audience - it's assumed
that from their point of view a "successful" technology is one that
shows up in their shop in mainstream applications. Per that
definition, SQL succeeded, OODBC not. Java yes Ada no. XML yes
SGML no. Et cetera.
5. The real point is that a bunch of things that ought to be predictors
for technology success (thus defined) don't work very well.
6. I claim that hitting an 80/20 point is a strong predictor of
success (positive and also negative in that technologies that don't
hit it usually don't succeed), that there are a few less strong
predictors (happy programmers, good early implementations, technical
elegance), and that military backing is a weakish negative
7. For the weaker predictors, I acknowledge counterexamples. I can't
think of one, either positive or negative, for the 80/20 criterion.
8. As regards 4GLs, that's probably my fuzziest test case. Someone
pointed out correctly that the really commercially important
outcome is VB. But anyone who lived through the 4GL hype in the 80s
(it basically amounted to "do away with programmers, and if you
can't, do away with 'if' statements") tends to look back with a very
jaundiced eye... and the Web seems to be giving poor old VB a tough
9. Simon suggested that XML was informed by HTML as much as by SGML.
Don't think so; XML is a very pure SGML subset and I honestly can't
think of any design components that came out of HTML. (Out of the
*Web*, sure, e.g. compulsory URIs for all external references).
10. Several people have hurt feelings over my suggestion that SGML was
not successful. Perhaps my opinion is influenced by having on
several occasions had to meet the payroll of a struggling SGML
company out of my own pocket. SGML "mattered" in the same sense
that Robert Johnson and the Velvet Underground matter to popular
music, but nobody bought their records and 99% of the programming
profession ignored SGML. Being important or mattering is not
equivalent to success in the sense that trade-conference audiences
think of it.
11. I've never heard a generalization about "SGML is better for X,
XML for Y" that holds much water. The most significant technical
difference is that for SGML the rules say you have to read and
believe the schema provided by the sender. Given that a lot
of successful SGML apps cheerfully broke that rule it's hard to
lend it too much weight.