Front page | perl.perl5.porters |
Postings from February 2018
Peter Rabbitson wrote: >The above statement is false. It still seems true to me. Perhaps you could explicate what other advantage the dual-location proposal has, relative to having signatures after all attributes. >Zefram's argument over the past several months is predicated on the idea that >signatures never existed in a stable form within the Perl5 ecosystem, and >didn't exist at all before 5.22. > >That idea is false (at best) My argument is in no way predicated on signatures not having existed before 5.22. The idea that they didn't exist certainly is false, and I'm acutely aware of it, due to my role in getting signatures into 5.20 and in arguing against them being damaged into their 5.22 form. I am mystified as to how you come to think that I would entertain that patently false idea. The idea that signatures have never been stable is true, unless you're playing games with the wording. The stable releases of Perl that have included signatures have all included explicit warnings, in documentation and at compile time, that signatures are experimental. They have thus never qualified as a stable feature, in our usual parlance. Perhaps you mean something else by "stable", but in that case you'll have to explain what you mean in order to further your argument. My argument doesn't entirely depend on the experimental status of signatures, though because they do have that status I have couched my argument in that context. Preceding attributes is the wrong place for signatures regardless of stability. Signatures are also much newer than attributes (and the :lvalue attribute in particular) and tacitly more experimental, regardless of formal status. The relevance of the formal stability status is only to provide a big explicit bias in the coherence vs stability tradeoff with which we are faced. It would still be reasonable to judge in favour of coherence without it. -zeframThread Previous | Thread Next