Mea culpa, mea culpa, mea maxima culpa.

On Fri, Mar 12, 2004 at 07:59:53AM -0800, Daniel Potter (Ars Analytica) wrote:
> I'd like to use a python class instances in _q_exports
> but can't figure out how to do the class _init_ at the
> appropriate times.
>
> Basically I want something like
>
> _q_exports = ['someclassinstance']
> someclassinstance = someclass('foo')

Oh, go ahead and call it class Taileater, I can stand it.  I can even
admit that I wrote that nonsense (well, not quite) on the TailEater
page at the wiki where it promised you that it "should be easy to use
it with a _q_exports list".  That was misleading without the
qualification that I was, then, working (and thinking) with plain old
CGI implementations.  As you've pointed out, it is *not* so easy to use
a TailEater object as the target of _q_exports in the general case.

The example was tossed together to add bulk and verisimiltude to an
otherwise drab and unconvincing, er, that is a rather short
description, and I never put enough thought into it to see the problem
with it in a persistent server.  Guess I'd better do something about
that...

FWIW, I *have* used TailEater (well, the trivially different version
that grew up to meet the need) with at least fastCGI (perhaps also
mod_python), and it works fine there when invoked from _q_lookup.  The
first_component initializer is probably another contributor to this
whole troublesome idea; that was an artifact of the original context
that has since become an unused relic, like an appendix, and it should
have been excised when TailEater was written up.

I doubt much of this is news to anyone at this point; I'm posting it
anyway in large part to remind myself to do something about these
problems with the TailEater writeup - this weekend with any luck.  I'll
just add a quick warning note now...

--
If there were a verb meaning "to believe falsely", it would not have
any significant first person, present indicative.  -- Wittgenstein