W3C home > Mailing lists > Public > www-rdf-interest@w3.org > August 2006

Re: RDF vs. relational databases

From: Jan Algermissen <algermissen1971@mac.com>
Date: Tue, 29 Aug 2006 17:12:24 +0200
Message-ID: <4544252.1156864344650.JavaMail.algermissen1971@mac.com>
To: Sören Auer <auer@informatik.uni-leipzig.de>
Cc: semantic_web@googlegroups.com, "revi s." <reviswami78@yahoo.com>, www-rdf-interest@w3.org

On Tuesday, August 29, 2006, at 04:42PM, Sören Auer <auer@informatik.uni-leipzig.de> wrote:

> From my point of view (next to the distribution aspect) flexibility is 
>a crucial difference between the RDF based Semantic Web representation 
>techniques and databases:

Yes, the need to organize instances into predefined, fixed types (tables) is detrimental to chageability (but provides the ability to optimize data access).

>With databases schema changes are very time consuming operations - the 
>whole repository and keys have to be reorganized. Triple stores on the 
>contrary don't distinguish between data changes and ontology schema 
>changes - both are finally just additions or deletions of triples. 

I have recently wondered how much flexibility you really gain. While there are no intensional types in the database (allowing you to shove in whatever triples you like and allowing any kind of reorganisation) the parts of a system that access the data still make quite a lot of assumptions on the sructure of the graph (about the existence of certain path patters, e.g. that a person has an address and that an address has a street). You cannot extract stuff without making such assumptions, but if you reorganize your store, the assumptions might become false.

While in the relational world such structural assumptions are part of the shared contract (the DB schema) they are implicit and only existing on the client side in the RDF world (and thus cannot be controlled).

IMO, it is of considerable importance, how you construct your queries (IOW: what assumtions they make about the RDF graph) and the goal should be to minimize the assumptions to really gain the flexibility RDF suggests.


>However, triple stores will probably never be able to compete with 
>optimized database schemas with respect to query speed. So if you need 
>high speed querying and you don't expect many schema changes use a 
>RDBMS, if you want to be very flexible with your schema/ontology use a 
>triple store. The RDF paradigm is also a bit more holistic in the sense 
>that everything from data, schema to metadata is encoded in triples, 
>while databases usually have different encoding techniques for each of 

Jan Algermissen                                               http://jalgermissen.com
Software Architect                                            http://www.tugboat.de
Received on Tuesday, 29 August 2006 15:12:58 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:44:58 UTC