At 05:34 AM 11/3/01, Dave Rolsky wrote:
>On Fri, 2 Nov 2001, Stephen Adkins wrote:
>
> > What should the supported Perl versions be for P5EE?
>
>People have already mentioned that we should think about what features we
>might want. We should also think about what bugs were fixed from 5.00503
>to 5.6.1.
>
>Also, 5.6.0 has some serious bugs that make doing things like throwing
>exceptions very problematic (its easy to trigger a bug that fscks the
>exception). Whatever we decide, I'd suggest that 5.6.0 should be
>explicitly unsupported.
1) I would second this. I hate 5.6.0.
However, with that said. It's important to note that many MANY ISPs use
5.6.0. When we re-released our software that had some problems with 5.6.0
to the open source community our support mails just shot through the roof
and we ended up fixing our software to work on 5.6.0 in those situations.
However, I think what we need to consider is that feature will be broken.
Will exceptions be used throughout P5EE? In that case, then I guess it is
necessary to do 5.6.0. But then we are explicitly making a design decision
to not let the "masses" use P5EE because the masses currently use 5.6.0
quite a bit.
2) About 5.5.3 vs 5.6.1
I would say it depends on the module set. There are many users legitimately
on 5.5.3 in larger production environments that don't have the time to
upgrade all their corporate servers to 5.6.1 if nothing ain't broke.
My belief is that most module sets should support 5.5.3 but it's possible
that for module sets that are standalone (eg the server side version of
POE) could run in a higher version separately. Or if it's not used often.
Auth, Authz will be used a lot I forsee. But a transaction engine might be
a special case, so I wouldn't mind that the transaction engine is forced to
use 5.6.1 but I would be saddened if the majority of users could not
benefit from a standard Auth API.
Later,
Gunther
