RE: [AF] relative URIs for attachments

I saw David recommend my version, presumably to get this wrapped up.  For 
the record, I can live with it either my way or Henrik's.  No need to 
check back.  Thanks.

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------







"Henrik Frystyk Nielsen" <henrikn@microsoft.com>
09/16/2002 07:55 PM

 
        To:     <noah_mendelsohn@us.ibm.com>
        cc:     "Herve Ruellan" <ruellan@crf.canon.fr>, <xml-dist-app@w3.org>
        Subject:        RE: [AF] relative URIs for attachments



>I think this looks good, thanks!

Thanks!

>Semi-unrelated to your changes, I'm a little unclear on whether there 
>really has to be just 1 URI to identify a part.

I don't think we say that "there can be only one" (tm) but that there is
"a URI" that identifies the part, see below for more.

>  First of all, 
>there is 
>the usual relative vs. absolute question, so at the reference 
>end at least 
>there are potentially multiples.

Well, as a relative URI only is meaningful with a known base, the
various combinations of relative URIs and base URIs don't count as
different URIs if the combination of the two is the same. For example, I
would say that the following all count as the same URI in the way we use
it:

                 base: "http://www.w3.org/foo/bar"
                 rel:  ""

                 base: "http://www.w3.org/foo/"
                 rel:  "bar"

                 base: "http://www.w3.org/"
                 rel:  "foo/bar"

>  Even beyond that, the web 
>architecture 
>generally allows multiple URIs for a resource.
>I'm afraid it does make the whole thing clumsier, but I'm reluctant to 
>appear to restrict the web architecture.  The case that particularly 
>concerns me is where a binding accesses resources in a lazy manner, 
>pulling them only when referenced.  In this case, should we really be 
>mandating that the part known as 
>http://www.w3.org/TR/2000/REC-xml-20001006 cannot also be 
>referenced in a 
>SOAP message as http://www.w3.org/TR/REC-xml (they are in fact 
>the same 
>resource on the web at this time.)  Saying that each part has 
>just one URI 
>rules out both of these references in a single envelope.  It's 
>a seemingly 
>obscure case, but I think it's better on balance to take the 
>trouble to 
>get it right.

But in this case, the reason why you know that there is a relationship
between the URI "http://www.w3.org/TR/2000/REC-xml-20001006" and
"http://www.w3.org/TR/REC-xml" is that you have read a statement making
this assertion and you happen to trust that assertion. In this
particular case, it's because the representation says so, but in fact
anybody could have made that claim. 

I don't think we in any way prohibit this: people are always free to
make assertions about URIs including statements about equality. I think
this is an integral part of the Web architecture that we inherit simply
by using URIs. For that reason, it is also not something that we can
outlaw. In other words, it is out of our hands to rule it in or out.
Therefore, all things being equal, I would prefer to stay with the
simpler model.

Henrik

Received on Monday, 16 September 2002 20:22:25 UTC