[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Deployment independent coding

On Mon, Jun 20, 2011 at 3:10 PM, Erik Huelsmann <..hidden..> wrote:
> On IRC, mst noted that we're using "very 1998 style" code and
> deployment-dependent style code, suggesting all 'new style' code
> should probably be written against PSGI
> [https://github.com/miyagawa/psgi-specs/blob/master/PSGI.pod].

Interesting document.
> Reading that document, I concluded we can't support that standard as
> little as we can support mod_perl at the moment: there's no way for a
> web app to indicate the script is to be run exactly once. And - with
> the built-in assumption of being a cgi script running only once - we
> can't support the multi-run paradigm for at least one other release
> (ie. we won't solve it in 1.3 yet).

We can't support it in 1.x.  The SL code breaks when you run it multiple times.

Not convinced it would be that hard to support for the new code since
the output code is centralized, but not sure that is really that
helpful and some SL code has been wired into the new output code, so
that is a recipe for a mess.

Once 1.3 hits RC status, however, I'd like to start an open discussion
on the future of 1.x and possible beginning work on 2.x.  This would
be a great time to consider this.
> So, my conclusion is that neither mod_perl nor PSGI or FastCGI are
> options as of yet.
> Thought I'd share that with you all. I'd like it to change though:
> we'll definitely need to have a look at which global values are being
> declared in order to eliminate them, if they're not truely global --
> in the sense that machine configurations such as paths may not really
> be problematic.
> Bye,
> Erik