On Sat, 2003-11-22 at 18:32, Greg Ward wrote:
> On 20 November 2003, Michael Watkins said:
> > all - i.e. the use of an object "database" such as ZODB.
>
> As a completely off-topic aside, I must strenuously object your use of
> quotation marks there -- it seems to me that you're implying either that
> object databases are not "real" databases, or that ZODB is not a "real"
> database.

Gosh, I consciously wasn't implying anything, other than perhaps my
crude understanding being that object databases are "different" than
relational databases.

Admittedly I've got a bit of relational database bias, having grown up
using them - Oracle, Sybase, Informix (shudder), SQL Server etc. My wife
is a db and design architect, and worked on one of the first Sybase
commercial projects ever in Canada.

So perhaps my subconscious took over there. Sorry! On the surface
however, my bias generally is being for anything that makes life easier.

And after writing something in Quixote with Postgres as a data store,
I'm now rewriting it completely and using ZODB (second attempt at this,
going much better this time). I love writing web apps with Quixote and
ZODB, its a nice fit for the kind of things I am doing.

But I admit to missing a query language more than ACID compliance,
although as I get more experience under my belt with ZODB its less of an
issue. Maybe being more comfortable with Python is part of that as well
-- back when I started, I could still whip up a complex multi table join
or a query of hierarchical data in SQL faster than I could in Python; I
bet that's changed now.

Why this is the second time around for me with ZODB - the first time, I
completed an application and had data integrity issues once it went
live, despite some amount of testing and load being put on the app.

After two weeks of troubles, some of which were no doubt my own lack of
experience with ZODB, and perhaps instability within ZODB or ZEO itself
-- I ended up converting the app to a Postgres backend.

But developing with an object db was so pleasurable that I'm back at it
a second time now, to figure out what went wrong.

> Au contraire, there's nothing particularly special about the relational
> model except that some database geeks in the 70s worked out a
> theoretical framework that means they can use Greek letters to prove
> theorems about databases.  Good for them.

I never understood most of what them were saying anyway...



--
Mike "didn't expect the spanish inquisition" Watkins
mw@mikewatkins.net


Scrubbing floors and emptying bedpans has as much dignity as the
Presidency.
                -- Richard Nixon