I have some simple questions about PIs since I'm beginning to try to implement
their processing.  

The current LANG draft  (961114) states:
The _Name_ is called the _PI target_.  It is normally the name of a declared
notation[1], identifying the application[2] to which it belongs.  The use of the
PI target "XML"[3] in any other way than those described in this 
specification[4] is a reportable error.

[0] I appreciate that the contents of PIs (apart from [3]) are of no formal
concern and that an author can write <?FOO ?> <?BAR BLORT?> as much as they
like.  However, since PIs are being extensively discussed as potential solutions
to some problems it seems useful to evolve some Generally Accepted Conventions.

[1].  Does this mean <!NOTATION...>?  If so, I can think of a lot of uses of
PIs which would be difficult to constrain within this.

[2].  Does application refer to the processing application, or to the 
"DTD + associated material and semantics".  Or something else?  I'm sorry to
sound confused but I don't have a background of SGML applications.

[3]. A number of PI-forms are being suggested in this group of the form
<?XML-stylesheet ...?> .  (This seems like a good way forward to me).  Does
the draft implicitly refer to PI targets of the form "XML-foo"?

[4] Does this refer to WD961114, or to the totality of the XML draft(s)?

It seems that there will be a hierarchy of PIs and that it will be useful to 
make sure they do not hiner interoperability and promote it if possible.
The hierarchy includes:

<?XML ...?>  So far this has Version, Encoding and RMD
<?XML-foo ...?> suggested as solutions to some problems
<?FOO ...?>

There seems to be a convention that PIs contain attribute-like components, and
this seems like a good idea.  Although it's outside the formal remit of the WG, 
clarification on xml-dev would help me, at least.


I hope there is no move to scrap PIs ...

In message <333F4E62.18F4@csclub.uwaterloo.ca> Paul Prescod writes:
> Processing instructions are another variety of "special comments." I do
> not think that we should remove processing instructions, and I think


Peter Murray-Rust, domestic net connection
Virtual School of Molecular Sciences