On Nov 5, 2009, at 2:09 PM, Jesse Luehrs wrote:
On Thu, Nov 05, 2009 at 01:10:19PM -0500, Stevan Little wrote:
At one point we felt it was sane to restrict what +foo attributes can
inherit. I think time has shown that we should probably just remove
this
restriction and let people have enough rope to shoot themselves in
the
foot.
There is a topic branch recently written that actually swaps the
approach we use, currently we have a list of "legal" options, the
branch
changes it to a list of "illegal" options. But honestly, at this
point I
think it would be just better to remove the restrictions entirely.
Care to whip up a branch that does that? I know there is some back-
compat issues with extensions that override the legal options method,
but I am sure they would be happy to take patches to just remove that
option. (Obviously we will need some kind of stub left for back-
compat
with older installs and which emits a deprecation warning).
I'm not entirely convinced here - at the very least, accessor names
shouldn't be allowed to be overridden, since doing so doesn't actually
make sense... the class would still be inheriting the superclass's
accessors. This should still be an error, I think. I can't think of a
reason why trigger would need to be excluded though.
Well what if you want to add a writer to a read-only attr? Variations
on that, etc etc etc.
- stevan
