> Can this alternate URI vary
> in any way? From installation to installation of a given XML application?
> From on-line to off-line operation? Over time in general?
Yes. Yes. Yes. Yes.
> In the face of
> whatever variation is allowed in the resolved URI of the external entity, what
> guarantees does the author of the external entity have about whether and what a
> relative URI in the external entity content resolves to (i.e., because the base
> varies, base+rel varies, leading to some uncertainty about nested content).
No guarantees whatever. This is a problem whenever a document can be retrieved
by more than one URL: it's not specific to public IDs.
> [D]oesn't the (non-annotated) spec already say that every ExternalID has to
> include a working URI?
Yes, except for ...
> In a NotationDecl, it is allowable to give only a Public Identifier without an
> accompanying System Identifier. Why does it make sense to offer this option
> for NotationDecl but not ExternalID?
Because the referent of the external id for a notation declaration is just an
explanation of the notation (in English or whatever), and it's not necessary to
fetch it to make use of the notation. It's most probably either ignored or
compared for equality, and public ids do either job just fine.
> It seems to me that for all the reasons people want URNs instead of URLs, we'd
> also like to have Public Identifiers that are simply not connected with those
> damn URLs (there are no really effective URNs available right now, so URI
> effectively equals URL at the present time).
URNs and FPIs (formal public ids) are the same thing, invented by two different
communities. In principle they could contain each other: there can be an fpi:
URN namespace and an +//URN FPI prefix.
-- John Cowan http://www.ccil.org/~cowan cowan@ccil.org You tollerday donsk? N. You tolkatiff scowegian? Nn. You spigotty anglease? Nnn. You phonio saxo? Nnnn. Clear all so! 'Tis a Jute.... (Finnegans Wake 16.5)