Re: OOP - a question about database access

The world rejoiced as "Bob Badour" <bbadour_at_golden.net> wrote:
> "Christopher Browne" <cbbrowne_at_acm.org> wrote in message
> news:bogm5j$1ef99t$2_at_ID-125932.news.uni-berlin.de...
>> I agree that it improves things to have a constraint applied in
>> multiple places, as opposed to _solely_ in the DBMS.
>
> Does it? Why?

Because in a heavy-duty distributed system, a lot of work takes place before data ever touches the database.

Supposing it's a "web application," it is good to constrain what is entered into web FORMs so that the DBMS doesn't have to roll back a lot of transactions because the user entered crap. So, for instance, it's probably a good thing for "upper tiers" to validate that dates are actually dates, countries are actually legitimate countries, and such, so that by the time the data gets shoved at the DBMS, there is _some_ likelihood that the data is at least quasi-legitimate.

>> After all, it's going to be really irritating to fill in a web page
>> with a bunch of data, and only discover that the values are
>> Seriously Messed Up when you try to COMMIT the transaction to the
>> database and discover it doesn't work out.
>
> It's also irritating when an error message prevents one from moving
> focus to fill in a field that would make the current field valid. I
> suggest it is even more irritating to discover that all of one's
> data is messed up because someone installed an old version of an
> application.

Oddly enough, I haven't seen this being a problem, given some forethought.

• With SAP R/3, the "GUI component" will refuse to connect to servers that are too new.
• If I mistakenly try to start up a PostgreSQL instance with software of the wrong version for the data, it'll abort immediately.
• <sarcasm> Many applications "run best" with Internet Exploder Version Something-or-Other, and may refuse to run otherwise </sarcasm>

>> The Right Way involves having some way of querying the database
>> about "Constraint Metadata", and then having some sort of code
>> generator that can transform that into Java/JavaScript/Perl/Python
>> as needed to support the upper layers of the system.
>
> Why do you assume this "code generator" is external to the dbms?

The code is external, and it would seem surprising for the code generator to be part of the very same program.

>> I think it's a real attractive idea to automate creating validation
>> rules that can be pushed upstream, but that doesn't make it easy...

> I suggest it is no harder for the dbms to enforce constraints on one
> computer than on another.

Foreign key constraints are an exception as they require going back to the central authority.

-- 
(format nil "~S_at_~S" "cbbrowne" "acm.org")
http://www3.sympatico.ca/cbbrowne/lisp.html
Security-wise, NT is a server with a "Kick me" sign taped to it.
-- Peter Gutmann in the Scary Devil Monastery