[cgiapp] Re: OT: Rose vs DBIx

[Aristotle Pagaltzis]

* Perrin Harkins <perrin at elem.com> [2007-12-26 23:10]:
> On Dec 26, 2007 1:43 PM, Aristotle Pagaltzis <pagaltzis at gmx.de> wrote:
>> But that can certainly be enough to make a qualitative
>> difference. After all, there's no difference in the
>> computations you can do with Perl or assembly language, just
>> the way you ask for them…
>
> Not a very apt analogy, in my opinion. From the examples I've
> seen, the difference in code required by these two database
> abstractions is not very big.

Well, my point is that just saying “the only difference is in
how, not what” is not a very apt assessment either. “Beware of
the Turning tar-pit” and all that. The interesting assement is,
how big is the difference in that how?

I didn’t say that their hows differ as much as Perl’s and
assembler’s, mind. They don’t. However, …

> Both are high-level APIs with a wide-range of query generation
> capabilities. Choices are likely to be driven by "feel" and
> personal preference.

… I do think that resultset chaining offers a real increase in
abstraction.

Sure – it doesn’t reduce the amount of code needed for any task
overmuch. What it does do is make it easier to build loosely
coupled pieces that each focus on a single task and which can be
freely combined and reused. Without chaining as part of the API,
more plumbing necessary to create that sort of architecture is
your own responsibility, so you’re less likely to build things
that way – you’re less likely to even *conceive* of building
things that way. It’s a bit of a Blub paradox situation, if not
quite nearly as stark.

Regards,
-- 
Aristotle Pagaltzis // <http://plasmasturm.org/>