Re: An object-oriented network DBMS from relational DBMS point of view
Date: Wed, 14 Mar 2007 18:04:00 GMT
>> On 12 Mar 2007 10:43:17 -0700, "Dmitry Shuklin" <shuklin_at_bk.ru> wrote:
>>>>That's not news, that an unconstrained description has constrained
>>>>grammers as a subset, that type 0 grammar has type 1,2,3 grammars as
>>>>"special cases", but the general case does not have some of the
>>>>properties of its simplified, special cases.
>>>Agree. But if some system allows to implement type 0 grammar then also
>>>it allow implement constrants and emulate type 1,2,3 grammars but not
>> Well, I don't know.
>> The value of an RDBMS is that it holds ALL the data.
>I don't necessarily agree with that. Some of the worst atrocities I have
>seen come from naive designers thinking normalization involves sticking
>every piece of text imaginable into some sort of lookup table.
>Then again, I would like to see a lot more managed data handled by a
>dbms. The key to that is extending the system to everywhere one needs to
I didn't expect the negative feedback on this, I find it interesting.
When I say, "all the data", what I mean is that the relational model
makes all the data in the database accessible homogenously, not that
it has to have my Macdonald's receipts listed somewhere! I do
understand the confusion, heterogenous data access is a long-time
favorite topic, especially with Bill Gates, going back to "NT5 Cairo"
and other misbegotten, lost initiatives. Anyway, when Dmitry proposes
islands of relational data in a larger storage engine, I'm just
pointing out that this misses out on one of the main motivators of
RDBMS. It may still be useful, but has no unifying concept.
>> language better than SQL), in the general direction of OO languges,
>> but it is the constraint on storage and the cannonical forms that make
>> RDBMS work, and just tacking on some swizzled spaghetti storage may
>> have its place, but it's really mixing apples and oranges, I don't see
>> that it enlightens either side.
>When you say "in the direction of OO languages", what specific features
>of OO languages do you desire?
Well, now that you ask, with a more detailed knowledge of SQL it seems ever-less important. As I said, the specialization of OO can be done with foreign keys, but it might be nice to have the alternative syntax. Might be nice to have methods local to tables, and resolving up some hierarchical arrangement of tables. Of course it's nice to have a full procedural language with SQL integrated, which PL/SQL has only had for twenty years, but having Java or .Net applets is nice and more OO. Whether it is good or bad to have object classes as opaque user-defined datatypes, I guess I can argue either way.
J. Received on Wed Mar 14 2007 - 19:04:00 CET