Re: PIFLE (was Re: Are PIs useful?)

I agree that requiring PI users to declare the namespace for PIs is a
good thing, however. If we are going to allow its use we should allow
its *safe* use. I just don't see that we need to standardize them beyond
that.

Rick Jelliffe wrote:
> If we make PIs hygenic, we needn't encourage people to avoid anything.
> They can stick PIs in for every specific target, without stuffing up their
> document's portability, as they see fit.

But the "sticking in PIs for every specific target" is exactly what they
are using SGML to get away from. That's why PIs should be discouraged.

> An example use for formal PIs is documents-in-progress. Editor settings,
> revision control numbers, etc, are all things that should be accessable
> in the grove, even if they are only to be used by the editor application
> & not the browser.  At the moment, they get plonked with a wink in
> comments or in elements.

It don't believe that a lack of formalism is stopping editors from doing
all of that now. I think that there are two things stopping them: #1.
Superstition, brought on by people like me who teach (preach?) that PIs
should be seldom used. I think that the examples that you site are
reasonable, but only SGML gurus write SGML editors so they know a thing
or two about parsing arbitrary syntax and should be able to get past the
superstition without help. #2. I've been told that some tools choke on
PIs. That's just poor programming. The semantics of a PI you don't
understand are simple to deal with: ignore it.

I don't think that formalizing PIs will fix either problem. I'm not even
sure I WANT to fix the first problem. I think it is GOOD that non-gurus
are afraid of PIs. I'm not out and out hostile to the idea of making
them easier to parse and work with, but I don't think it is necessary.

 Paul Prescod

Received on Wednesday, 14 May 1997 23:38:26 UTC