Re: [entity-resolution-comment] Some queries on the XML Catalogs spec

[Paul Grosso <pgrosso@arbortext.com>]

At 22:00 2003 04 14 +0100, Richard Tobin wrote:
>> >Is the system identifier input to the resolver the system identifier
>> >as it appears in the document, or the result of absolutizing that?
>> >Presumably the latter, but it doesn't seem to say that in section 7.1.
>
>> It should be the former, though I understand some processors such as
>> SAX make this difficult.  But I believe it should definitely be the former.  
>> Otherwise, how could you make an entry that worked, for example, for:
>
>>   <!DOCTYPE book SYSTEM "docbook.dtd">
>
>I had assumed that you would always use a public identifier in the case
>of "well-known" doctypes, but I don't have much experience of actually
>using catalogs!
>
>The thing that worries me about using the relative URI is the
>possibility of "URI-capture".  There must be a million (ok, a
>thousand) people who have made up their own DTDs called "book.dtd" or
>"address.dtd".

You are right that PUBLIC makes more sense in many cases, but
I know that SYSTEM of the relative path makes sense in some cases.

I'm not too worried about URI capture.  By judicious use of a list
of catalog files (with more "private/local" ones coming first in
the list), I don't see this as a problem.

In Arbortext's distributed TR 9401 catalog, the last line "system
default" catalog includes:

  DOCTYPE html "xxxx"

where xxxx points to the installed HTML doctype.  This is serious
"doctype capture" where any document with a doctype decl with "html"
as the document element is going to use this entry.  And we've never
had any serious problems.

>> Personally, I'd think it would be fine to treat the XML catalog as
>> well-formed XML and then do any error checking you wish--including
>> validating it against a built-in DTD.
>
>When I looked at the DTD the thing that struck me was that you will
>often need to apply it just to get the defaulted (fixed) namespace
>declaration on the root element.  Or do people usually put it in
>explicitly?


I don't have an answer here.

paul