On Thu, 08 Sep 2016 07:43:02 +0000, Jack Stouffer wrote: > On Wednesday, 7 September 2016 at 19:01:23 UTC, Piotrek wrote: >> Almost every "standard" evolves (e.g. USB, 3GPP, etc) and are subject >> to change in subsequent releases. Stopping the progress is not a case >> in good standardization process. > > When I say "a good candidate for standardization", what I mean is a > standardization of an API and module design, not a standardization in > the traditional sense. > > It doesn't matter that a standard like HTTP2 will have a new version > (e.g. 2.1), what matters is the way in which the programmer interacts > with it and how that API is designed. If there's no clear answer, e.g. > urllib2 vs. requests, then that probably shouldn't be included in the > standard library. Continuing with the urllib2 example, how many people > do you suppose use urllib2 over requests, which is the most popular > Python library by far? Despite this, the Python team is stuck > maintaining urllib2.
Your example shows that hasty API design can be undesirable. Nobody's objecting to that. The current pattern in D is: * propose a module * get feedback on the newsgroup * implement it and put it on dub in a std.experimental namespace * get feedback from people actually using your work * make PR * get feedback from people who already maintain Phobos * it's now in Phobos This pattern was/is being followed for, if I recall, ndslice, checkedint, xml2, and logging. Possibly a couple others. Do you think that process is insufficient? Or do you think that it's not useful to push these libraries into Phobos?
