Re: System URIs

[John Cowan <cowan@mercury.ccil.org>]

Rob Lugt scripsit:

> John, just to make this absolutely clear, are you saying that the following
> entity declarations:-
> 
> <!ENTITY etude1 SYSTEM "http://example.org/étude">
> <!ENTITY etude2 SYSTEM "http://example.org/%C3%A9tude">
> 
> ...should each match all of the following catalog entries:-
> 
> <system systemId="http://example.org/%C3%A9tude" uri="etude.ent"/>
> <system systemId="http://example.org/étude" uri="etude.ent"/>
> <delegateSystem systemIdStartString="http://example.org/%C3"
> catalog="example.cat"/>
> 
> If this is what you are proposing

It is.  Since the two forms are strictly synonymous (technically, only the
%-escaped one is a URI, the other is an extension to URI syntax), they
should match the same set of candidates.

> It needs
> to point out that both sides need to be %-escaped before a comparison is
> made.  If the XML Processor has already %-escaped the system identifier,
> then the catalog resolver only has to %-escape the systemId and
> systemIdStartString.

Just so.

> In the case of URIs, I don't think there is anything in XML 1.0 or
> Namespaces describing %-escaping.

XML Rec 4.2.2 as corrected by Erratum 7, and as posted by me earlier, does
in fact specify that %-escaping be done according to the method of RFC 2396
whenever a non-URI character appears in a system identifier.  This includes
both non-ASCII characters and URI-invalid ASCII characters as well.

>  So the catalog resolver will have to
> %-escape both sides

Only if the resolver is construed as part of the XML processor.  If it is
construed as part of the application (and there is warrant for both cases),
the %-escaping will already have been done when the resolver gets the URI.

It's not clear to me whether this TC or the XML Core WG will have to nail
this down, but *someone* will.

-- 
John Cowan                                   cowan@ccil.org
One art/there is/no less/no more/All things/to do/with sparks/galore
	--Douglas Hofstadter