I am thinking about the general tools and the cultivated impression that
PIs are somehow bad.  There is the XML specification then there are the
common practices built over a bed of experience, opinion, and naivete

PIs are just out of band information.  That can be useful.  If one
doesn't accept that, one doesn't face the reality of global
communication networks and languages evolving dynamically.  Not all
information is meant to be processed by all processors at first or at
all.  If an application language designer (eg, the SOAP designers) wants
to restrict PIs, they can.  If others find them useful, they can use
them.  If two groups conflict, the XML specification doesn't care
explicitly. The market works out the winners and losers.   If a group
decides to use that opportunistically, that's fine.

So bits like DOCTYPEs, DTDs, PIs, etc., we have to live with because
they work.  Namespaces?  One doesn't have to use them but the case for
using them is better than the case for ignoring them.   MusicML is
likely an isolate: it doesn't have to aggregate other XML languages, so
avoiding the namespace and using simpler means such as PIs makes sense.


Conservation of Power:  Never use an amplifier where a filter will do.

len


From: Tatu Saloranta [mailto:cowtowncoder@y...] 

--- "Bullard, Claude L (Len)"
<len.bullard@i...> wrote:

..
> As to PIs, the fact that so many XML processors
> don't handle 
> them says the XML specification designers' job was

I'm curious: are you thinking in general terms of XML
processing applications, or strictly about XML
processing tools (parsers, xslt/xquery tools)? I
thought that latter must handle PIs (whereas support
for comments was more optional), and at least
regarding parsers, they must be passed to handling
app/tool.

That obviously doesn't help with folks who re-define
what they think XML should mean (Soap, Xmpp) and omit
pieces they don't like.