On Fri, Aug 30, 2013 at 12:43 PM, Aristotle Pagaltzis <pagaltzis@gmx.de>wrote: > * Ricardo Signes <perl.p5p@rjbs.manxome.org> [2013-08-30 14:45]: > > You know, I have been so consumed with faith in this feature for years > > that I felt it could just drop right in. I will admit that I may be > > biased. If there is consensus that all or part of this feature should > > be experimental (which seems reasonable) then okay. > > I don’t foresee problems with postfix deref, but then I didn’t foresee > the edge cases in the maximally minimal signature proposal either, which > started to turn up once it was really scrutinized. The interaction > between the `each @array` and `each $autoderefed` features weren’t > foreseen either. Etc etc. Everyone so far has been sure that their pet > feature would work great. > > P5p’s track record for collective piecemeal language design is… middling > to put it mildly. And I don’t just mean the 5.10 disaster, it goes back > way farther – think pseudohashes, say. In fact every non-trivial feature > I can think back to turned out to have major caveats to work out – if it > wasn’t entirely ill-conceived. > __SUB__? s///r? The callchecker? \&CORE::? Not saying that you don't have a point, but the picture is nowhere near as grim as you've depicted it. (It's too early to tell for my personal favorite, /(?[])/) > > We really don’t have much grounds for confidence. > > So I was kinda assuming that now that we have the mechanism, *all* major > new features would have to make a round as an experimental feature, as > a matter of policy, to avoid the kind of clusterfuck/embarrassment of > the likes of smartmatch. > +1Thread Previous | Thread Next