Dept: re-opening a can of worms :-) Two straw-man proposals have been put forward to two separate groups for a PI that would permit an XML document instance to name a schema or schemas against which it is (supposed to be) valid.
In April 2005 I wrote a document that at the time was circulated only among the TEI Council members: http://www.tei-c.org/Drafts/edw89.xml?style=printable In July 2005 MURATA Makoto wrote a document that at the time was circulated to the new Relax NG user's mailing list: http://www.asahi-net.or.jp/~eb2m-mrt/hidden/spec.html In both cases the documents are just proposals of their authors, not position papers of the organizations they sometimes represent. In both cases the syntax of a PI that an XML document may use to point to a schema is described. In both cases the intent is that the PI itself is optional and repeatable, and that while processors may want to make use of the information, they are not required to. This issue fostered some discussion on the rng-users list back in summer 2005, with some objecting to the idea of a schema-pointing PI (Tommie Usdin I recall was among them). TEI P5 needs to be finished up soon, and it would be a good idea, I think, to have this particular issue settled well before it is published. Because TEI is so customizable, it is quite helpful for the user if an instance can point to its schema. In general, the TEI would be as happy or happier if someone else (read OASIS) came up with the specification for this process, so that TEI P5 could refer to it. My instinct is that a PI is the right place for this information, but I'm not even sure that is a given. I don't think there is really that much work to be done in this area to come up with a unified proposal. Does any one else think it is worth any effort? Is this something that comes under the jurisdiction of the OASIS Relax NG Technical Committee?
