Re: The trouble with absolutizing...

"David E. Cleary" wrote:
> 
> > The trouble is simply that there are thousands of documents Out There
> > which identify namespace names based on identical string matching,
> > that will become broken if the rules are changed retroactively.
> 
> Which is why we can't retroactively forbid using relative URIs too. So the
> question becomes want can we do that is between forbidding relative URIs and
> requiring absolutizing.

I guess that in reality you are asking what we can do to tell the
difference between ambigous interfaces. Generally speaking, you would
choose between something like 1) duplicate 2) extend 3) fix.

If you want to preserve the functionality of the thousands of documents
allready out there you MUST NOT do 3.

If you don't care about history you SHOULD do 3.

If you want to preserve the functionality and extension mechanism you
SHOULD do 1.

My opinion on this is somewhat similar to what others have expressed,
but I'll express it here anyway.

Assuming that we do need namespaces, it's just a matter of what we want
out of it. My requirement spec is that they should enable both global
and local use. It should also be possible to prepend or append (or both)
them without altering their meaning to enable a higher namespace
resolution at will and at need. Personaly I would not require
"relativliness" explicitly since it could be accomplished by tools
prepending or appending the namespace value.
I don't possess a vision of a semantic web [i can even admit that only
guess the true meaning of it] but if this can be accomplished with
URI's, fine with me, if not I like java packaging. Anyway, it's not a
requirement of mine and I expect it to be possible to accomplish some
other way.
I would really like to see an example of what URI's would look like if
you would have both Semantic Webs and unlimited resolutions. In the end
I think it's just a matter of which micro-parser the tool-smiths have to
build/use, but maybe that's a way too simple point of view.


Cheers,
/Niclas

Received on Tuesday, 16 May 2000 15:16:12 UTC