Re: Storing query language in relations

From: Mikito Harakiri <mikharakiri_at_yahoo.com>
Date: 3 Oct 2002 11:50:00 -0700
Message-ID: <bdf69bdf.0210031050.4addb7df_at_posting.google.com>


"Tibor Karaszi" <tibor_not_pressed_ham_.karaszi_at_cornerstone.se> wrote in message news:<NzKm9.306$MV.13285_at_newsc.telia.net>...
> > > Are there DBMSs where you can "create" new tables, views, indexes,
> > > constraints, etc. just by inserting into the appropriate system
> > > tables? So the DDL becomes redundant?
> >
> > I also thought about it ~yr ago: it is certainly a nice idea. DBA have
> > to learn only catalog schema, no stupid DDL.
>
> Perhaps I'm missing stuff at beginning of this thread, but isn't DDL
> actually to be seen as a *good* thing?
>
> By inserting stuff directly into catalog, you make that "DDL" highly product
> specific. And DDL it is, although in the form of INSERT statements. All
> current RDBMS's that I know of have their own layout of the system catalog.
>
> ANSI has defined the INFORMATION_SCHEMA view, but these are but views of the
> actual system tables. ANSI SQL has defined the DEFINTION_SCHEMA tables, but
> this is, even in the ANSI papers, to be seen as an imaginary set of system
> tables, for the sake of defining the INFORMATION_SCHEMA views (my own
> words). I know of no product that has actually implemented the
> DEFINITION_SCHEMA tables (although I wouldn't be awfully surprised if Ocelot
> has).

Why DEFINTION_SCHEMA tables are necessary in the vendor's implementation? A vendor just need to implement INFORMATION_SCHEMA views upon their (proprietory) catalog. Then, a user just updates INFORMATION_SCHEMA views, which then triggers updates in the underlying catalog.  

> Also, at least MS SQL Server has to do other stuff when you do DDL. For
> example:
> First, you have several tables. You have one catalog table for the table
> existence. And another for column existence. And possible also for the
> physical characteristics (where to start a scan, for instance).
> Second, initial space allocation probably has to take place. Agreed, this
> could be handled by a constraint of some sort, and above could be taken care
> of by a deferred constraint. IMO, this wouldn't buy us much, though.
Received on Thu Oct 03 2002 - 20:50:00 CEST

Original text of this message