xml-dev - RE: SAX2 Lexical Handler Suggestions

[Eric Bohlman <ebohlman@netcom.com>]

On Thu, 13 Jul 2000, Kay Michael wrote:

> A general rule: if your application needs to know where the entity
> references are, then you are not using entities the way they were designed
> to be used.

Or you're writing an editor or editor-like application in which, as has
been pointed out before, you need something more than a "simple" API.  The
main characteristic of such applications is that they apply
transformations to human-maintained documents.  If, for example, a
document author defines some entities as "symbolic constants" he or she
would not be a happy camper if the process of editing the document, or
have having bibliographic references looked up, normalized out all those
symbolic constants, any more than a C programmer would be if running his
code through a profiler replaced the preprocessor-defined constants in his
source.  Similarly, if a document is physically divided into multiple
storage units, a transformation shouldn't automatically lump everything
into a single storage unit.

Such applications are relatively rare, but they do exist.  In designing a
truly generic parser API (which, of course, SAX isn't *by design*),
supporting them is probably 90% of the work for 10% of the applications.  
Of course, the original poster was using physical entity structure to
convey logical information, which is a no-no (similar to the one in WML
where a literal dollar sign has a different syntactic meaning from a
numeric character reference to a dollar sign).