Re: rdf-utils 0.2: support for diff and patch

Reto Bachmann-Gmür wrote:

> not sure if the diff/patch functionality of rdf-utils was 
> misunderstood, it is of course independent of serialization 
> idiosyncrasy (i.e. independent of the order of statements and b-node 
> labels). The advantage of the concept of rdf-molecules[1] over the 
> concept of minimum self contained graph is that it takes (inverse) 
> functional property into account.
>
While using some IFP might readily fit some use case, I see a number of 
issues on the other hand with dividing graphs into molecules rather than 
MSG.  Especially if you want to do RDF Synchronization. From the paper 
is see that starting with 6 triples (page 9) you end up with more 
molecules than triples. In fact it seem exponential with respect to the 
statements attached to blank nodes "islands".
MSGs on the other hand are a unique, non overlapping decomposition of an 
rdf graph, no redundancy here.
It is also argued that this is a list of possible graphs that once 
combined have the same semantics.

True, but during the transfer (that is in patch form) they do not have 
the original meant semantic, it seems to me that "molecules" are smaller 
(potentially) than the smallest particle of information which conveys 
the intended meaning, example:
one might have said "I know X and X has mailbox foo@foome.com X has blue 
hair and X is 9 feet tall X is a creation of my fantasy, disregard it" 
you basically split this into molecules , each molecules carries a 
partial view of what the original person meant.
.. some molecules will not have the fundamental part saying that X "is a 
creation of my fantasy".
Would the author agree that such molecules carry his intended meaning 
individually? would the author agree that just some of these molecules 
be added with its name to an external DB?

On the other hand adding IFP is straightforward extension to definition 
of  MSG (it is in fact mentioned in the published papers ), but then 
again if you start minding owl arent there a lot of things that can be 
done  ("same as" smushing etc, way more complex reasonings based on 
cardinality etc..)?

>
> 1. http://www.ksl.stanford.edu/people/pp/papers/Ding_ISWC_2005.pdf
>
Its a bit sad that these papers dont reference MSGs (first paper 
published in [1]), but i see not even URIQA is mentioned, strange.

Wondering, what happens if the same incremental path (e.g. add a my 
previous example "I know X and X has mailbox foo@foome.com X has blue 
hair and X is 9 feet tall X is a creation of my fantasy, disregard it" ) 
is applied twice to the same graph (note, no IFPs are used here.. assume 
"has mailbox" is not an IFP in this example) ? do you get X1 and X2  or 
a single X node?

Giovanni

[1] 
http://semedia.deit.univpm.it/submissions/ISWC2004_workshop_p2p/RDFGROWth_workshopISWC2004.pdf

> Giovanni Tummarello schrieb:
>
>>
>> If model RSync is what you're looking for, the code in the 
>> RDFContextTool lib ( http://www.dbin.org/RDFContextTools.php but the 
>> code to use is really the one on the CVS) allows building a 
>> fantastically efficent one for RDF graphs (RDF RSync or R2Sync). 
>> Experimental results look really nice and the procedure is based 
>> exclusively on RDF theory (and the Minimum Self contained Graph 
>> theory  [1], also used for digitally signign such fragments), not on 
>> the way some DB decides to store the graph (e.g. some DB specific 
>> de/serialization idiosyncrasy).
>> I just need to find some time after the DBin release to get it out in 
>> in the public, but if someone is in urgent need of traffic efficient 
>> syncronization, please msg me I might be able to help right away  :-)
>> Giovanni
>>
>> [1] G. Tummarello, C. Morbidoni, P. Puliti, F. Piazza, "RDF signing 
>> supporting resource centric requests" Proceedings of the Poster 
>> track, ESWC 2005.
>> http://semedia.deit.univpm.it/submissions/ESWC2005_Poster/ESWC2005_signignRDF.pdf 
>>
>>
>> Danny Ayers wrote:
>>
>>> On 1/7/06, Reto Bachmann-Gmür <reto@gmuer.ch> wrote:
>>>
>>>  
>>>
>>>> rdf-utils 0.2: support for diff and patch Reto Bachmann-Gmür 
>>>> 2006-01-07
>>>> 24:26
>>>>   
>>>
>>>
>>>
>>> Wonderful!
>>>
>>> Coincidentally I just ran into this, looks like you just covered 2/5 :
>>>
>>> http://www.daml.org/2001/04/iow/mit/index.htm.old
>>> Semantic Web Development
>>> Intent of Work
>>> 23 March, 2001
>>>
>>> ·         Parsing and regeneration of DAML to/from a data store
>>>
>>> ·         Persistent RDF/DAML data repository module
>>>
>>> ·         In-memory RDF/DAML data store
>>>
>>> ·         Repository difference (delta) calculation module
>>>
>>> ·         Repository synchronization module
>>>
>>> ·         Basic Inference rule processor
>>>
>>>
>>> Cheers,
>>> Danny.
>>>
>>> -- 
>>>
>>> http://dannyayers.com
>>>
>>>
>>>  
>>>
>>
>>
>>
>
>

Received on Sunday, 8 January 2006 20:05:11 UTC