[Python-Dev] RELEASED Python 3.0 final (original) (raw)

Guido van Rossum guido at python.org
Sat Dec 6 19:16:21 CET 2008

On Fri, Dec 5, 2008 at 10:03 PM, <glyph at divmod.com> wrote:

The best thing for 3.0 adoption would be a 3.0 "welcoming committee". A group of hackers wandering from one popular open source library to another, writing patches for 3.x compatibility issues. There must be lots of people who care about 3.x adoption, and this is probably the most effective way they can reach that goal.

Each time I am going to fix a 3.0 compatibility issue, I have a choice: I can either make Twisted itself better (add features, fix bugs), or I can keep Twisted exactly the same but do lots of work so it will work on 3.0. It seems pretty clear to me that, to the extent that I have time for Twisted, fixing bugs in the HTTP implementation would be a better deal than puzzling through a megabyte of diffs generated by 2to3, trying to understand where it went wrong, and how. This doesn't mean I'm "sitting on my hands". It just means I have better things to be doing with my hands. (To be precise, 1054 better things to do, re: Twisted. Add in the Divmod projects and it's more like 3000.) Of course the distant threat of an unmaintained 2.x series is enough to motivate me to push a little in this direction, but it doesn't make me happy about it. I think this is exactly what the marketing effort around 3.0 needs to be doing: making a positive case for library and application authors to spend time to update to 3.x. This is a lot of work, and many (I might even say most) of us need a lot of cajoling. Free patches are a good incentive :).

This is a really good idea. I hope and expect that the information and tools available for porting to 3.0 will dramatically improve over the next half year or so (hopefully the situation is a lot less gloomy already by the time we meet again at PyCon). The porting list that was just created also sounds like a step in the right direction.

I do think that in many cases some support from the regular maintainers of a library would be needed -- for example if you (in particular) were to express a negative attitude towards porting Twisted to 3.0 (I'm not saying that you do, it's just a hypothetical that would apply to any "BDFL" for any sizable library) then this would discourage others from trying to contribute. OTOH if you made a branch available where you check in the results of running 2to3 over Twisted, with instructions for people to contribute fixes, that would be great -- at almost no cost to you! (Assuming you can get someone else to work on merging trunk improvements into that branch.) Remember the open source mantra -- reap the benefit of all those eyeballs!

-- --Guido van Rossum (home page: http://www.python.org/~guido/)

More information about the Python-Dev mailing list