PI and entity specs in XML draft of 14-Nov-96

Section 2.6 says that the name which follows "<?" in an XML processing
instruction, the "PI target", is "normally the name of a declared
notation, identifying the application to which it belongs."  What
exactly is the force of "normally" here--- is it just saying "we hope
this is how people will choose to use this feature"?  The only actual
requirement on a processor seems to be the special behavior for PI
target XML.

This isn't really a complaint, but a request for a bit more
forthrightness: perhaps "normally" (or some other term like
"typically") could be included among the definitions of 1.4 with the
explanation "XML is often used in this manner, but conforming data and
software need not behave as described."  There are other places where
describing "the usual thing" would not be a bad idea---notably in the
case of the usual way of handling external graphics files with entity
references.  SGML types know to use entities and notations; HTML types
are likely to be puzzled at the presence of these features and could
use at least a nudge towards their use.

John Lavagnino
Martin Hensel Corporation and Brown University