Lists Home |
Date Index |
[apologizes if this appears twice, this is a resend after a mail problem]
Over the weekend I was reading through the XML Catalog specification 
and toying with the implementation of a catalog resolver available
through the Apache xml-commons project . I was surprised about the way
that system and public identifiers are handled, and thought I'd check in
here to see whether I'm misinterpreting the spec (likely!) or could find
some rationale for the behaviour.
Specifically, it would seem that System Identifiers are the preferred way
of accessing a catalog, and Public Identifiers play second fiddle. Granted
the spec does include a "preferred" mode for accessing the catalog but
according to Section 7.1.2 of the spec if a system identifier can be matched
in the catalog then that will always be used; an accompanying public
identifier is only used as a fallback if the system id doesn't match
(or is unavailable) and the prefer setting is "public". So public ids are
only useful if the system identifier isn't available. Given that you
have to provide both in XML, public ids seem pretty redundant.
Toying with the command-line resolver application shows behaviour consistent
with this reading.
Everything I've read up to now set me up to expect the following: that
Public Ids would be preferred and that System Ids would be the fallback.
What am I missing?