W3C home > Mailing lists > Public > semantic-web@w3.org > January 2008

Re: plural vs singular properties (a proposal)

From: Renato Golin <renato@ebi.ac.uk>
Date: Mon, 07 Jan 2008 11:36:41 +0000
Message-ID: <47820EC9.2050406@ebi.ac.uk>
To: editor@content-wire.com
CC: tim.glover@bt.com, semantic-web@w3.org

editor@content-wire.com wrote:
> However, it is still the case that RDF cannot capture the relational
> model - there is no way in RDF to specify that a property has a unique
> value.

Hi all,

I'm getting a bit late in the discussion but let me give my impressions...

Uniqueness cannot be defined from the data itself, there must be
something external (like indexes) that guarantees uniqueness.

Because any RDF is mergeable to any other RDF (and that's a very
important feature of RDF to be lost), there is no central control
blocking users to insert duplicated values.


> provided nobody disagrees, could you (or others) comment on the
> implications of this?
> how do we create/support data integrity in RDF?

To impose unique indexes to RDF would be to revoke automatic
mergeability and that's a very high price, I wouldn't go in that way.

There is, however, very complicated ways of automatically merging any
kind of data to any other kind of data to avoid duplication in a set but
because with RDF the database is the whole world there is no way of
assuring uniqueness in the global sense.

So you can have uniqueness in a set of semantic data, you can merge
other data and still guarantee uniqueness but you can't guarantee global
uniqueness as relational databases can because the scope of the later is
a single table while for the former is all relational data in the world.

The problem does not stops there, this is uniqueness in the triplet
level, which means that you still could have:

foo is bar
foo means baz
foo est bal
foo sameAs rab
...

in the same RDF even when using triple indexes. It's easy to show that
even having ontologies mapping "is sameAs est" and blocking the later
insert you still can't guarantee that all similar terms will be mapped
in ontologies at all.

Unless there is some control over the data and how it's written there is
no way whatsoever of assuring consistency and uniqueness in a data set.

my pence...

cheers,
--renato

-- 
Reclaim your digital rights, eliminate DRM, learn more at
http://www.defectivebydesign.org/what_is_drm
Received on Monday, 7 January 2008 11:37:10 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:20 GMT