Re: public/catalog [was: Re: ERB Decisions of March 26th]

On Fri, 28 Mar 1997 18:40:55 -0500 Jon Bosak said:
>  Could those who claim to be interested in seeing public
>identifiers implemented in XML please comment on Paul's proposal?

I think it's close to the Right Thing and thank Paul for his
draftsmanship.  Having a rule for catalogs that says all XML processors
have to recognize PUBLIC entries and may skip the rest seems to me very
close to the &mntdv;, so I think we should build it in and declare

About finding the catalog, I favor a slightly different solution:

  1 first, check for a local catalog (location and method of
    specifying location are part of software installation, outside
    scope of XML -- but XML processors should always allow for a
    local catalog)
  2 if the local catalog doesn't exist or doesn't have a matching
    entry, then check the default catalog, which is
      a  the catalog named inside the document in an appropriate
         element or PI or whatever, if the document
         uses an architectural form or DTD known to the processor
      b  the catalog found by looking at the relative url "./catalog"

That is, provide a meaningful default even in the absence of a specific
instruction, just as RFC 1738 does (or at least the Fielding draft does)
in the case of Base URLs, and furthermore use basically the same rules
as RFC 1738:  if there's specific instruction in the document, then
follow it.  If not, use the location of the object itself, and look
'nearby'.  The fact that Panorama works this way I regard as a clue that
(a) it works, and (b) it will be familiar to many potential early

I think having a standard way to find catalogs is critical to
interoperability of XML processors, so I do want to specify it, though
Paul is correct that it is in some sense independent of using catalogs.


(Oh, yes, for those who were puzzled, there's an implicit declaration of

  <!ENTITY mntdv 'minimum needed to declare victory'>

before this text starts.)