[xquery-talk] Re: The State of Native XML databases
john.snelson at oracle.com
Mon Aug 20 18:54:53 PDT 2007
Thank you for your considered response.
Jeff Dexter wrote:
> I've previously had the argument with Ilya on the collection vs.
> document notion, and while I do disagree this is a relational concept, it's
> important to understand that his use case does not work well with the
> concept of collections. He's dealing with very large, single documents
> representing the collaboration point and he requires a high degree of
> concurrent access to them. We tried melding the collection concept to this,
> but in his case he's constrained by a standard schema and so shredding into
> sets of smaller documents would not work. Hence the document is the
> database, in this use case.
I agree there are use cases that don't split into many XML documents
well, but one document is still a collection.
> In regard to XML Schema, leaving all discussion on the failing and
> complexities of the language aside, once you do have the darn things written
> they can be quite helpful in authoring XQueries, typing your collections and
> other activities so they are germane to the original discussion of the state
> of native XML databases. In my anecdotal experience for every person that
> squirms when I mention XML Schema in their application there's another who
> already has a few dozen cooked up and is raring to use the
Some type of schema can of course be useful in different scenarios, but
Ilya's suggestion that "ddl=xml schema" does not suit everyone.
> Finally, "enterprise" and "native" are indeed bandied about
> altogether too much, but again Ilya brings up a good point in the discussion
> of the state of native XML databases, which is how individual databases and
> more general XML platforms handle things like locking/concurrency,
> transactions, etc. which are feature very relevant to large scale
> applications, and again, an interesting aspect to use in contrasting the
> various implementations.
Agreed - although it seems more pertinent to talk about the use cases
solved rather than the implementation details. Node level locking is,
after all, merely a means by which an engineer hopes to achieve better
concurrency - which is the real use case.
I have no reason to suggest that TigerLogic does not fulfill Ilya's use
cases, nor that it shouldn't be considered along side other XML
databases. I just object to his implications of a religious and blanket
dismissal of other XML databases and other use cases.
More information about the talk