Lists Home |
Date Index |
In the SGML On The Web WG/SIG debate about features
to toss, some of us wanted to get rid of PEs. The
counter argument was that some DTDs reached a level
of complexity that they became unmaintainable without
PEs. Others argued that at that point, it might be
better to use multiple DTDs and refactor the documents
themselves, or just get used to cut and paste.
They are useful, but I have tried to avoid them whenever
possible mainly because they make a DTD painful to
read. To understand a DTD, I always have to
replace these with the actual productions.
Those who argue they are like object-encapsulation,
a design tool, a modularity feature, and so forth,
have a weaker point than I like for keeping a feature
so prone to errors.
From: Michael Kay [mailto:email@example.com]
> I am very wary of proposing errata to fix this kind of thing,
> because it is so easy to produce unintended consequences.
This is an admission that the whole business of parameter entity references
is so fragile that no-one dares touch it in case the whole edifice comes
tumbling down. This is not a good foundation for the e-commerce
infrastructure of the 21st century. When this happens to code that I'm
responsible for, I throw it away and start again.