Re: XML catalog draft

Paul Grosso wrote.
>I don't understand the automatic translation of which you speak.  It
>sounds like you're talking about translating system ids into public
>ids, but that's "backwards." 

The question is "how do I create a public ID in a catalogue without asking
the author to do more than point to the thing he wants added to the
catalog?". What I am suggesting is a way wherby an author can supply in his
system set-up a public identifier source specification and then just point
to a file identifier and tick a box that says Add To Catalog?

> That may be something of interest for an
>authoring system, but an XML application still needs to map those
>public ids back into system ids to do anything.  How is that to be

The catalog still needs a map from the public ID to the system ID, but this
is easy. In addition, if it can't find a system ID, it can check the
contents of the public ID to see if it can identify a "word" that has a # in
it, and try to use that as a system ID.

> If through a catalog, then you'll need to set up the catalog
>that maps the public ids to system ids, and I don't see why you can't
>do that for valid public ids as easily as for invalid ones.  If you are
>suggesting that your invalid public ids "contain" the system id and can
>therefore be mapped back into the corresponding system id without a
>catalog, that sounds like a system-specific, non-interoperable
>mapping--which basically leads me back to saying your invalid public
>ids might as well be system ids.

What I'm looking for is the minimal amount of work that a system would have
to do to create a catalog entry for a linked document. There have been a
number of comments in the link thread about needing to be able to create
XML-links by point and click techniques. The same must be true for catalog
entries. Somewhere in betweeen lies the management problem. Once the initial
point and click operation creates a catalog entry it must be possible at a
later date to change the system ID without altering the public ID, even if
this means that the URL component of both no longer match. (The fact that
they no longer match is a good indicator that the system identifier is a
better bet than the original name, and could provide a useful check on which
links should be validated first: those that still contain their original
address could, for example, be given a higher priority in any checking process.)

We need to understand better the relationship between catalog entries, link
groups and link definitions before we finalize things. I just want us to
keep the idea that having an # in the public ID could be useful in providing
us with a minimalist method of creating catalog entries.

Martin Bryan, The SGML Centre, Churchdown, Glos. GL3 2PU, UK 
Phone/Fax: +44 1452 714029   WWW home page: http://www.u-net.com/~sgml/