Reporting as requested, there are two things our projects
continue to need xml2 for, and a third which we conceptually
need xml2 for:
(1) We need some way to do xslt processing.
(2) We need the equivalent of xml_is_well_formed(text), e.g.
a documented way to call XMLPARSE speculatively and get a
NULL or be able to efficiently catch an error, etc.
(3) My third point is a bit subtle. The xml2 package is
clearly just a set of operations on text. The PostgreSQL
xml datatype might RIGHT NOW be internally represented as
text, but I don't think we're supposed to think of it that
way.
The official XML datatype is free to evolve into one of
the more sophisticated XML storage and processing formats,
at the expense of a bit more overhead at XMLPARSE time.
The advantage of the xml2 package is that it says upfront:
this is just a way to treat text as xml. It is useful
precisely when one is doing a mixture of text and xml
processing steps on data, and perhaps also when generating
short-lived xml values.
(4) In conclusion, I hope that PostgreSQL will keep xml2
or something similar even when the XSLT and speculative parsing
issues have been addressed.
Thanks for your consideration,
_Greg
J. Greg Davidson