RE: Validation, Versioning and RDF [Was Re: WSDL WG request for adding multiple version extensibility into Schema 1.1]

David Booth wrote:
> IMO, this is the main reason
>relational databases won the database wars 20 years ago when they were
> introduced: you can easily add new tables (i.e., new relations) to a
> relational database without breaking existing application code.  
> However,
> if you add new data to an XML schema for a tree-structured data format (for
> example by inserting siblings or by inserting new levels of hierarchy) you
> are much more likely to break existing application code.  It is possible to
> avoid, but it's harder than in the relational world.

i do like the analogy with relational (v) hierarchical databases, 
but one view of an RDBMS is as a loosely connected graph with 
joins between columns and views. 

I agree, the success of relational databases was in no small part due 
to their making  touchless versioning possible - the query language 
/by default/ ignores new columns and tables, and missing data can
be easily recognised and given a default value.

Paul

Received on Wednesday, 3 March 2004 12:05:20 UTC