Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Usenet -> comp.databases.theory -> Re: Sourcing Metadata for Database Independence

Re: Sourcing Metadata for Database Independence

From: Kenneth Downs <knode.wants.this_at_see.sigblock>
Date: Sat, 05 Mar 2005 20:37:15 -0500
Message-ID: <>

Nolan wrote:

> Kenneth Downs <> wrote:

>> You could instead commit 1000 hours to the engine/generator (put that

> one
>> unusual programmer on it, the guy who gets things done when left

> alone),
>> and devote another 4000 hours to having non-programmers spec out the
>> metadata, and feed their specs into the engine and try zillions of
>> combinations until they get what they like.
>> The manager who believes that it will take 10,000 hours is tied

> forever to
>> the results of his decision: a crawling horror that must be

> maintained
>> until Judgement Day.  Worse, the next job he does will be done the

> same
>> way, and he will put millstones around his neck one by one.  The

> toolmaker
>> on the other hand, is tied only to a flexible tool that he can re-use

> on
>> the next job.
>> I'd rather be tied to the tool :)

> Sorry for chiming in late on this discussion.
> Has anyone seen such a tool currently in use?

Does my office count? :)

Our reference implementation builds PostgreSQL databases and creates web apps that run identically in the big three browsers (IE, Mozilla, Opera), with the web glue running in Apache/PHP.

> This approach sounds like the CASE tools of the '90's, which have not
> provided the promise they held out for a more flexible, re-usable
> applications development environment.
> Nolan Madson

From what I understand, the primary difference between Andromeda and its closest relatives is our strict emphasis on being able to *completely* specify an application in data. This is actually required by the logic layed out in my original post, in which data is seen as the lasting asset. AFAIK, nobody else goes that far, and IMHO that's why they become bogged down as the "exceptions" eventually dwarf systematic stuff and we all know where that leads.

When it comes to any particular coding language, our attitude is "this too shall pass." Our complete reference implementation is actually astonishingly small (or at least I think so) at less than 10,000 lines of code for a complete 3-tier system incorporating security, automation, constraints, and so forth. The inputs vary according to the count of tables and columns in the system, and the outputs go the same way, but the builder and libraries stay small.

Rather than CASE or RAD, I would term Andromeda an "ASA" tool, for "Automated Software Assembly" because that is what it actually does. It maps features in the data dictionary to trigger snippets, CREATE TABLE commands and the like, and then provides a copy of the data dictionary to the web layer, which does the same thing.

Kenneth Downs
Secure Data Software, Inc.
Received on Sat Mar 05 2005 - 19:37:15 CST

Original text of this message