[Prev][Next][Index][Thread]

Re: the return of the Public Identifier Question



[Donning the hat of a simpleton]

1.
What does a "system identifier" really mean in this day and age anyway?
"foo.bar"
on my "local" system can be "harry.belafonte" on a machine in Madagasgar thanks
to distributed systems stuff like NFS, Unix Symbolic Links etc. Is the 
attempt at absolute/relative addressing implied by PUBLIC/SYSTEM not thwarted
by all this stuff?

2.
Has the problem of Universally Named Thingies Carved In Stone For All Time been
solved anywhere, in any system? If so, should we not use it? If it has not been
solved why kill ourselves trying to crack an uncracked genus of nut?

3.
Here is a suggested solution that is, as the saying goes, "simple, elegant and 
wrong".

Make all XML entity names URLs. Allow (we cannot stop!) arbitary levels of 
indirection to be implemented at OS level to map a URL to something 
else.

Define a simple URL to URL mapping CATALOG scheme as part of the XML spec.

Put in the spec. a cautionary tale about how eclectic URLs are an
accident that has already happend. I.e. 
Anyone implementing a WEB site would do well to avoid publishing URLs like
"/x/y/z/w/b/foo.xml" and instead use "/marketing/foo.xml" having mapped
/marketing to "/x/y/z/w/b" in their server....

Sean