Re: The fable of DEMETRIUS, CONSTRAINTICUS, and AUTOMATICUS

From: Laconic2 <laconic2_at_comcast.net>
Date: Fri, 22 Oct 2004 08:45:08 -0400
Message-ID: <GrWdnaxkX5hDneTcRVn-rg_at_comcast.com>


"Marshall Spight" <mspight_at_dnai.com> wrote in message news:EPZdd.293253$D%.166424_at_attbi_s51...
> "Laconic2" <laconic2_at_comcast.net> wrote in message
news:DJadnbrwrt5efOrcRVn-jw_at_comcast.com...
> >
> > BTW, one of the features of DEC Rdb/VMS that I used to like was the
> > "$RMU/EXTRACT" feature.
> >
> > This command would examine the system tables of a database and create an
> > entire CREATE script that would
> > create the database, and all the CREATE commands needed to set up
database
> > objects, schemas, and schema objects inside it. The create script was
SQL
> > DDL (DEC Rdb dialect). So "reverse engineering" an Rdb database was
like
> > shooting fish in a barrel, if you knew how to RTFM.
>
> Now, if we have a catalog, why can't we just dump the catalog
> and insert *that?* Metadata is still data.
>
> It seems to me that DDL must at all times be considered simply
> as a shortcut for doing DML on the catalog.
>
Not quite.

When CREATE TABLE gets executed, two things get done:

First, the table gets created.
Second, the catalog gets updated to reflect the existence of the table.

Each of these steps is inconsistent without the other.

If I simply did DML on the catalog, the first step would not be done. Unless of course, there were some kind of trigger mechanism on the catalog tables to invoke the system functions that make the actual structures agree with the catalog. Received on Fri Oct 22 2004 - 14:45:08 CEST

Original text of this message