[
Lists Home |
Date Index |
Thread Index
]
Peter Hunsberger wrote:
> On 7/19/05, Ronald Bourret <rpbourret@rpbourret.com> wrote:
>
>>Will the addition of XML data types to relational databases help this
>>problem, at least for data-centric schemas? It's a bit of an force-fit,
>>but it seems that mixed content could be mapped to and accessed from an
>>XML column when the meaning of that content is a single word, even an
>>annotated one.
>
> If you're indexing in any way (walking a Blob or shredding) then you
> need some form of schema support...
Actually, virtually all native XML databases can index arbitrary XML. In
some cases, this is inherent in the internal structure of the database.
For example, NeoCore stores data in hash tables of the paths to that
data. Thus, the index and the data storage are the same. In other cases,
databases create more conventional indexes (such as B-trees) as
necessary, rather than simply in response to a create index command.
Whether this is always done or the user has the choice to limit what is
indexed depends on the database.
>>This solution won't work for data binding until programming languages
>>can handle XML data as a first-order type.
>
> As I pointed out yesterday, there's more-or-less a direct mapping
> between XML mixed content "annotations" and Aspect oriented
> programming annotations.
Good point, although it seems this would only work for the kinds of
data-centric annotations Rick was talking about, as opposed to generic
mixed content, such as is found in XHTML or DocBook.
> Building a mixed content data structure
> isn't the problem, at this point I don't see how to generally use
> schema driven parsing to populate such a structure. It would be
> straight forward populating such a beast directly via a hard coded SAX
> parser. The issue is, can you automate the building of the parser
> without some form of meta-meta description? I can't see how, and
> existing Schema languages don't appear to be up to the task.
Neither can I. It's worth noting, however, that virtually all data
binding products are schema-driven, so this might not be much of an issue.
-- Ron
|
