Lists Home |
Date Index |
Uche Ogbuji writes:
> This is why I have long called for a generic constraints system
> rather than a monolithic type system for specs such as XPath. Not
> that generic constraint annotations will solve all such problems,
> as you go on to point out. In the end, program safety and
> correctness is not something that can be automated by computer,
> despite the remarkable claims of strong data typing advocates.
This kind of data typing is extremely convenient if you're using XML
as an object serialization or an RPC format; it might also help you
optimimize data storage a little (but not as much as people expect).
In fact, I used to be a big advocate of adding something like
'xml:type' containing a URI (sorry, Simon et alii) identifying a data
type. After a few years and a lot of big XML projects, I don't really
see the point any more. If there is a real need for typing, people
can gather around something informal and prove it, i.e.
> XML has proven invaluable as an undecorated bridge to the
> ultimately responsible human, which I argue is its greatest
> strength. I'm not sure how or when this goodness came under attack
> from considerations of "wizards". I guess, as Simon St.Laurent
> said, it was probably right after XML took off and the recalcitrant
> mainstream programming types realized they had not quite signed up
> to what they expected.
I don't think there are (many) people who want deliberately to
obfuscate XML. The problem is that people believe it desirable to
write specs before there's a proven need. That leads good, talented
people into all kinds of wild guesses and bad decisions (sort of like
Soviet-style central economic planning, which many Western countries
imitated to varying degrees in the 1970's and early 1980's). This
problem is not unique to the XML community.
All the best,
David Megginson, email@example.com, http://www.megginson.com/