Re: RDF comments from the ages

Brian McBride <> wrote:

>> Spec treats email addresses as literals, not URIs
>> (probably should be added to errata)
> Whilst email addresses can be represented as URI's, I don't see why it
> is an error when they are not.

It's not really an error, but it'd be nice to fix.

>> First known instance of Qname->URI problems:
> This issue is already present as:

First known instance, etc. means I found an older thread and felt that it
should be linked in. So you better link all those in or I'll continue to
badger you for a long time. ;-)

>> The first known instance of the set or bag question
>> DanBri cites the use of set terminology as evidence towards a set:
>> EricP says that in a multi-user system, it should be a bag (sorta):
>> Bleed-over from XML-L:
> has been generalised to cover the absence of any formal description of
> RDF graphs in the spec.

I think this might be more appropriate to go under:

>> conformance problem with xmlns I hadn't seen before:
> I don't see a problem here.  RDF has no unique attribute requirement.

It's not really a problem. XMLNS has a unique attribute requirement, RDF
doesn't... it's just a little odd.

>> rdf:ID should create isDefinedBy triples

Did you add this one? I feel it is important. Perhaps under:

>> Melnik: I proved that the most general RDF model does not exist.
> I think Sergey's claim is that he proved it does exist.

Whoops, sorry. You're right.

>> We need a standard way to talk about the source of triples:
>> http://www-db.Stanford.EDU/~stefan/updates.html
> I don't see a specification issue here.  Reification can be used to
> model the source of statements.

I think the issue is more that there is no way to talk about the source,
while continuing to assert it. Somewhat related to:

>> How do you assert something, but give the resulting triples a URI?
>> (i.e. assert and reify at the same time)
>> This makes time difficult to represent in RDF
>> Gabe: That's not true -- use rdf:ID

Which I think should be combined into a new issue, as they are essential for
doing any large-scale agreement systems in RDF.

> This is really a web architecture issue, however, I think we'd all be
> better off we had a clearer formal model and terminology for URI's and
> what they name, perhaps something similar to what the topic maps folks
> have done.
> I'm going to seek some advice on this one.

I agree, and I think Dan hinted towards something on the subject. If nothing
else, we should at least keep track of all the issues and discussion so that
they can be turned over to whoever takes care of this.

I personally think this really needs to be resolved, at least sometime, by
the SWeb activity.

> The above are all about the nature of URI's.  See above.

I know. I grouped all together for you. ;-)

>> Do anonymous resources have a URI?
>> DanBri: URI is just one bit of info about a resource
>> Two kinds of URIs (private and public):
>> Melnik: Agreed-upon URIs for anonymous resources are essential
>> DanBri: You can't do that -- you need more info:
>> DanBri: anonymity is not a property of a resource, but of a mention:
>> Melnik: Good summary of problem/solution:
> Added reference to Sergey's summary.

What about all the others? -- they're all very informative (or at least my
summaries are). ;-)

>> Can't define acceptable properties for a class:
> I'm not sure I see an issue here, given that anyone can say anything
> about anything.

The issue is that in many OO languages (and databases) you constrain classes
by stating the properties that they have, like:

    :Car x:hasProperties :color, :price, ...

RDFS allows you to constrain properties, as in:

    dc:title :domain rdfs:Literal

but you can't do the same (and more common) think for objects. Since for
just about everything else RDFS' expressive power is about equal for classes
and properties, I'd at least like an explanation why this is different.

>> Issue for Brian: Can you make the issue IDs links to their URIs? Like:
>> Issue <a
>> href="http:...#rdfms-uri-for-graph">rdfms-uri-for-graph</a>: blah blah
>> blah...

Any response on this?

I'll have a bunch more complaints when due course (and CVS commits) are over
and the updated version is on the site.

Aaron Swartz <>| 
  <>   |   <>
AIM: JediOfPi | ICQ: 33158237|  the future of news, today

Received on Tuesday, 13 March 2001 20:18:52 UTC