[Numpy-discussion] Meta: help, devel and stackoverflow

Matthew Brett matthew.brett at gmail.com
Thu Jun 28 18:03:39 EDT 2012


On Thu, Jun 28, 2012 at 2:57 PM, Fernando Perez <fperez.net at gmail.com> wrote:
> On Thu, Jun 28, 2012 at 2:07 PM, Matthew Brett <matthew.brett at gmail.com> wrote:
>> I see that sympy, for example, has only one mailing list, and that
>> works extremely well.  I'd be interested to hear from the Cython and
>> IPython guys as to whether they feel the user / devel split has helped
>> or hurt.  Ferando? Dag?
> There's evidence that projects can work successfully in either mode
> (single/dual lists), so I don't think this is a completely clear-cut
> question with a 'right' and a 'wrong' answer.  What matters most is
> finding for each project and community what works best, and I think
> the main factor should be how truly disjoint are the topics and
> typical threads of the two lists.
> Before talking about IPython, we can consider Python itself, where
> there's a very clear division between the general and dev lists, and
> even the dev list has been recently split with a new 'ideas' list
> where more exploratory threads can take place, so that -dev can remain
> 100% focused on active, concrete development work on the main Python
> repo.  And that strong separation of lists (which python-dev enforces
> strictly by calmly but firmly redirecting threads to other lists as
> soon as they seem off-topic for the narrow python-dev focus), seems to
> work pretty well for them.
> As far as IPython, I personally do prefer the separated lists, and I
> think it works quite well for us.  IPython is a project often used by
> python beginners for simple learning of basic programming, and they
> just want to know how to tab-complete or how to get plots to run in
> non-blocking mode.  Our -dev list is relatively high-traffic and with
> a weird mix of topics, given the rather eclectic nature of IPython: we
> have qt discussions, parallel computing, low-level networking/zeromq,
> javascript/web issues, protocol API threads, etc.  All that can be
> overwhelming for novices (though obviously one hopes that novices
> would gradually learn from that and become interested in being
> developers).
> I think this is how I'd summarize it:
> - having two lists is friendlier to beginners, as it gives them an
> environment in which to ask questions that they may feel more
> comfortable in, because the level of the discussions tends to be not
> as complex as what happens in a -dev list.
> - but the cost it has is that it insulates users a bit more from the
> development ideas, perhaps lowering the likelihood that they will
> catch on to the development conversations and dig deeper into the
> project.
> My cartoon view of it would be:
> a. novice person | user list  || dev list
> b. novice person || combined list
> where the | bars indicate 'barriers': in (a), a novice has a low
> barrier to become a good user, but a higher barrier to transfer into
> developer.  With (b), there is no clear barrier to becoming a
> developer, but it's more intimidating for new users to join.
> I have heard (but I only have anecdotal evidence) of users saying that
> they feel more comfortable asking questions in user-only lists because
> of the level of the discussion, and that they can read all messages
> and learn something without having to filter threads that are way over
> their heads.
> Long answer, I know... But in short, I'm happy having two lists for
> IPython: I prefer to have the first transition (gaining active users)
> to be the easiest to make, because I think once users have become
> confident, the cost of digging deeper into development is actually
> pretty low.
> But I'm sure other projects can and have successfully made the opposite choice.

Fernando - you told me a week or so ago that you'd come across a blog
post or similar advocating a single list - do you remember the



More information about the NumPy-Discussion mailing list