Why database modules are incomplete

[Paul Boddie]

Dave Cole <djc at object-craft.com.au> wrote in message news:<m3g0c4hstf.fsf at vole.object-craft.com.au>...
> >>>>> "Edward" == Edward Wilson <web2ed at yahoo.com> writes:
> 
> Edward> Dave: The way I see it, nobody is going to use a module "trial
> Edward> and error" in the workplace, especially if they suspect it may
> Edward> be incomplete.
> 
> If they suspect that it is incomplete - maybe they should spend a few
> hours to see if that suspicion is well founded.  In most cases there
> are only one or two modules to try.

What could be interesting would be some kind of common test suite for
database modules. It would give module developers something to aim for
(for compliance with the DB-API), and module users something to use to
verify how good a module is for their purposes. mxODBC has something
like this for sanity checking, as well as for investigating certain
aspects of type binding.

> Edward> They are going to use a proven module, or hack one up in their
> Edward> shop's native tounge.
> 
> And the one hacked up locally will automatically be better than the
> one which already exists which you could not spare the time to
> evaluate...

Sadly, in-house code often gets hacked up for database access, but the
reasons can be persuasive - if you don't have a high turnover of
staff, you can at least have a group of people who are very familiar
with the code. It seems slightly ridiculous, though, to duplicate
effort because people would rather write code than read someone
else's.

[...]

> Edward> lending our time as testers.  My personal feeling is that if
> Edward> ten or twenty folks would jump in and help you out, in a year
> Edward> (working part-time) Python could have a diamond hardened
> Edward> Sybase module.
> 
> You are 100% correct here.  If only people were willing to try the
> module.  For all I know there are more than 20 people who are using
> the module for serious work.  Noone ever tells me, so how can I claim
> that I have a proven module?  As I said before - there have been over
> 180 unique IP addresses which have downloaded the Sybase-0.2x module.

Perhaps it works sufficiently well for the purposes of the users who
downloaded it. Nevertheless, sure-fire ways to elicit feedback would
be invaluable for module developers. Do any of these Sourceforge
surveys ever get used? (Or are people too worn down by having too many
meaningless "Who is your favourite actress?"/"What kind of geek are
you?" surveys in their face whilst surfing?)

> Edward> Python is way cooler than PL/SQL or any other language used
> Edward> for writing db apps for that matter.
> 
> Which is exactly why I wrote the module - to avoid using ISQL or Perl.

These are certainly reasons that I won't contest. ;-)

Paul