Re: Modeling question...

From: paul c <toledobythesea_at_oohay.ac>
Date: Fri, 14 Nov 2008 22:48:28 GMT
Message-ID: <0fnTk.385$jr4.293_at_edtnps82>


Bob Badour wrote:
> paul c wrote:
>

>> Bob Badour wrote:
>> ...
>>
>>> You were talking about physical performance. Thus, the key concept is 
>>> not logical independence via views but physical independence via 
>>> whatever structures the dbms has available to physically store 
>>> various representations of the data.
>>
>> That's the way I think Fabian Pascal saw it too.  Regardless of 
>> whether one thinks the shortest logical path is the only logical path, 
>> it has annoyed me no end that the (few) dbms's I've used didn't allow 
>> me to insert my own logic in between their storage level and the user 
>> view.

>
> Well, they do and they don't. Many dbms products allow for various
> indexing and clustering schemes. Far more options are needed and the
> scope of the physical dbms should be much broader than it is now.
> Perhaps with the advent of multi-core chips and cloud computing dbmses
> will start expanding that scope.

Yes, but the products I've seen allowed index specification only with respect to a single "table" without any option to make it into two tables. So I'd say that is a different sort of physical option. Still, it is kind of ironic that one of the first forms of data clustering, combining two "tables" into one (as opposed to index clustering) came out in a hierachical dbms (might have been IMS, I forget), whereas the relational ones at least circa fifteen years ago hadn't bothered. In fact, none, if any, had any notion of predicate locks, which I think was the one real lasting good idea of Jim Gray's. Received on Fri Nov 14 2008 - 23:48:28 CET

Original text of this message