xml-dev - Re: Using processing instructions as transaction semantics

[Lars Marius Garshol <larsga@garshol.priv.no>]

* Chris Barney
| 
| What does the group think of using processing instructions as
| transaction semantics vs. using elements
 
Processing instructions sound like the wrong thing to me. They are
mainly for things that do not fit in any schema/DTD, but rather cut
across all schemas/DTDs. The xml-stylesheet PI is a typical example.
Another (although rather ugly) example might be the PI used by a
certain SGML editor to indicate where the cursor was.

The name processing instruction may have lead you astray here, since
it does not refer to transactions or anything like them. Instead,
processing instructions were originally intended to be used in SGML
documents (where all presentation had been stripped in favour of
structure) for those cases where presentational hints to the
formatting software were needed.

With time they have come to be used as a sort of generic ignorable
syntax extension, and that is what they should really be regarded as.
If you can use elements or attributes, don't use processing
instructions. 

| Does either approach have noticeable advantages/disadvantages?

PIs have the disadvantage that they live outside all schemas and DTDs,
which means that you cannot constrain their use except through the use
of BNF and prose.

--Lars M.


***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************