Re: out-of-band transformation information

Further on this thread:

At SCO we used a set of REL values which were employed by the browser.
There was no value associated with the profile attribute and no namespace
declaration because there was no XML or namespaces then.

The browser had knowledge of the REL values. Our entire UNIX, X11, Motif
and all other documentation (which would fill half of a normal office bookcase)
was encoded using one of Dan's early HTML DTDs from 1993 along with a
few REL values. The browser had a small set of toolbar icons that were related
to the link REL values.

The browser asserted authority over the meaning of link elements in those 
documents.
And we documented the fact so that anyone using the scohelp browser could take
advantage of this service that the browser offered.

There is no saying that future browser (or user-agent) makers will not 
initiate their own
policies to interpret the meaning of REL values and META scheme/name/value 
triples.
That is, although an author or designer neglects to include a GRDDL 
Transformation link,
an agent remains at liberty to infer a non-GRDDL Transformation link 
according to its own policies.

Dan designed GRDDL so that authors and namespace designers could make 
assertions
that are essentially binding contracts between data-providers and 
data-consumers.
There is no room for third parties to be involved in these contracts, 
morally or legally.

The role of a third-party transformation might at best be paraphrasing and 
at worst
completely distorting the meaning of the data. In any case, provenance 
becomes uncertain
and the chain of evidence is broken. That's not to dispute the potential 
usefulness of
3rd party transformations, but they are not in keeping with Dan's original 
intent to
enable forming binding transformation contracts among willing participants.

Just another 2 bits chipped into the pot.

Murray