On Sat, Jan 3, 2015 at 1:27 PM, Joachim Durchholz <j...@durchholz.org> wrote: > Am 03.01.2015 um 13:02 schrieb Sergey B Kirpichev: >> >> On Sat, Jan 03, 2015 at 12:27:41PM +0100, Joachim Durchholz wrote: >>>> >>>> Lets hope that mpmath will not break backward compatibility. >>> >>> >>> I wouldn't want to rely on hope. >> >> >> In fact, this is a sane assumption about library development. > > > I strongly disagree about this being a sane assumption. > > It may be sane for libraries that already have a varied set of consumers, > which is proof that they have committed to a stable API. Even there, I have > seen libraries make horrifying decisions.
Fredrik said that he will try to keep the API stable, so it will work out fine. > >>> Let's stick to the one version we test for during a release. >> >> >> Why? This is unnecessary > > > Why? For the reasons I gave. > These reasons are unrefuted, you just snip them and claim the opposite - > what do you suppose that I do, bow to your superior intellect? > Sarcasm aside, you're not going to convince me by "because I say so". > >> and only add difficulties for packagers. > > Well, that's a problem for the packagers to solve. > > I'd say they should be helped if they present concrete problems and > cooperate for a solution. For example, if they want to allow additional > versions, and we find the testing overhead exceeds what we can do on Travis, > it would be appropriate if we could use their test farm. Or maybe they can > convince the Travis guys to give us faster testing. > > If the packagers can't help us, then maybe their problems aren't that > relevant to the distribution that they're working for, and they'd need to > solve the problems themselves. > I'm sympathetic with their situation, I'm not going to let their priorities > override ours, which has reliability as its primary purpose. > > If this is going to be a problem, then maybe unbundling mpmath was the wrong > decision. I think we can solve technical problems like testing on Travis. I agree with you that pip is a horrible way to install and maintain software, but fortunately there are now pretty good package source managers like Conda or Hashdist, so you don't need to use it. Neither of these two existed when we first wanted to decouple mpmath, but today there solutions exist and while not perfect, they are improving and are very usable. Ondrej -- You received this message because you are subscribed to the Google Groups "sympy" group. To unsubscribe from this group and stop receiving emails from it, send an email to sympy+unsubscr...@googlegroups.com. To post to this group, send email to sympy@googlegroups.com. Visit this group at http://groups.google.com/group/sympy. To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CADDwiVC2V4YH9D2aEw8zmrjSCQCq0eY2%2Byd1nzngmOQ_Z_qLOw%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
