Re: lack of support for claims regarding Concise Bounded Descriptions (see MSGs)

From: Giovanni Tummarello <giovanni@wup.it>
Subject: Re: lack of support for claims regarding Concise Bounded Descriptions (see MSGs)
Date: Mon, 06 Jun 2005 15:20:26 +0200

> 
> >I don't know if this would solve the issue.    It depends on your wording.
> >
> >
> >Consider the graph
> >	 ex:a ex:b _:a .
> >	 _:a ex:c ex:d .
> >The MSG of the first triple includes the MSG of the second
> >triple, which includes the MSG of the first triple.  So if something is in
> >the MSG then it is in, but just what is in?
> >
> Actually the MSG of the first is the same [as] the msg of the second.  

Well, yes, this is wwhat makes sense?

> if one is included in the other then they're the same.

Yes, but just what are they?

> i'll look into wording it better and let you know. Giving the concept i 
> believe is clear what wording change would suggest?
> Does it count if i show the code as definition? ;-) it is really simple.

Code doesn't really cut it, unless it is very high level.

You could probably use something like Patrick's definition of CBD.  Adding
something like "the smallest ... such that" might also work.

A definition that I believe works is:

The MSG of a statement s in an RDF graph G, MSG(s,G), is 
the smallest subset of G such that 
1/ s is an element of MSG(s,G); and
2/ if s' is in MSG(s,G) and x is a blank node that is the subject or object
   of s' then MSG(s,G) includes all statements in G that have x as a
   subject or object. 

[...]

> >>that peers dont all have to know the same thing, one might be interested 
> >>justin madonna another just in that song .. the'll have some overlapping 
> >>information but not all of course.
> >
> >Suppose that you could just transfer the statements one at a time, with
> >bnode identities retained.  This would allow transmittal of the entire
> >graph.
> >
> If bnodes can have identities that persist across models.. well they're 
> not bnodes by definitions.

Ahh, but who says that we are working in the *semantic* domain?  If you are
willing to "look inside the black box" then lots of other stuff is possible.

[...]

> Giovanni

Peter F. Patel-Schneider
Bell Labs Research

Received on Monday, 6 June 2005 13:48:37 UTC