[Pandas-dev] Mailing list for Python data analytics ecosystem developers?

Matthew Rocklin mrocklin at gmail.com
Sat Dec 29 17:58:36 EST 2018 

> I don't see what is wrong with using e-mail.

There were some issues raised before:

   1.
*I'm slightly concerned that a broad ranging e-mail list that encompasses
   all of PyData would get noisy.  For example I can imagine lengthy
   conversations on visualization or probabalistic programming that, while I
   find important, would likely want to take a pass on.  Having a service that
   includes tags and subscription to those tags may have value. *
   2. *E-mail list archives tend to collect dust.  If we're doing
   long-range cross-project coordination then those conversations might have
   long term value.  We might want to cross reference them, upvote them,
   subscribe to them, and so on.*

And also some benefits of discourse raised by Nathaniel which might be
turned around to be interpreted as concerns with e-mail.











*My impression so far is that discourse takes a bit of adjustmentcompared
to mailing lists, but it has a lot of valuable features likemulti-quoting,
markdown (code blocks, links, ...), solid moderationtools (e.g. if a
discussion diverges you can retroactively split partsof it out into a new
topic), polls (these were incredibly useful fortaking the temperature of
the community during the governancediscussions), ability to reply to
messages that were posted before youjoined the list, configurable
notifications (email me everything /email me when a new topic is created /
email me a summary weekly /...), ...*

> It is public, archival, and append-only. GitHub issues are non-archival
and comments can be edited or deleted.

That's certainly true of GitHub issues.  I suspect that it's also true of
Discourse (though I'd have to go through the docs to make sure that it
wasn't possible to turn it off).  From my perspective the (in)ability to
edit or delete comments isn't a big deal.  I'm not particularly concerned
with people modifying history in a nefarious way.  Though perhaps my
viewpoint here is naive.  I haven't yet run into this issue in our
community.

I think that the biggest benefit to using an e-mail list is that it's a
well known technology with a low barrier to adoption.

I anticipate two likely failure modes for e-mail and discourse respectively:

   - EMail: conversation is too diffuse so that people sign up, get bored
   listening to things that don't interest them, and then stop notifications.
   The pydata mailing list ends up being used by small subsets of the
   community, but not the community as a whole.
   - Discourse: it's too new/unknown so that no one signs up and it doesn't
   reach critical mass.  (this seems to be happening with Jupyter's discourse
   today?)

There are lots of other pros and cons to each, obviously, but those two
outcomes are, I think, the most troublesome.

On Thu, Dec 27, 2018 at 8:13 AM Wes McKinney <wesmckinn at gmail.com> wrote:

> Having dev.pydata.org sounds fine to me.
>
> I don't see what is wrong with using e-mail. It is public, archival,
> and append-only. GitHub issues are non-archival and comments can be
> edited or deleted.
>
> On Thu, Dec 27, 2018 at 6:26 AM Andy Ray Terrel <andy.terrel at gmail.com>
> wrote:
> >
> > I would recommend we set up a site dev.pydata.org that tells the folks
> where conversations are happening. While mailing lists are great we might
> consider just having a github issue tracker set up for cross ecosystem bugs
> or initiatives. I was planning on decommisionning the numfocus discourse
> and zulip server as they didn't really have great use. Chris Holdgraf
> suggested using Topic Box, but user based pricing isn't a great option for
> open source development.
> >
> > Anywho, both dask and pandas are part of the NumFOCUS projects ecosystem
> so I'm happy to set anything up for ya'll.
> >
> > -- Andy
> >
> > On Wed, Dec 26, 2018 at 10:35 PM Wes McKinney <wesmckinn at gmail.com>
> wrote:
> >>
> >> @Andy
> >>
> >> pydata at googlegroups.com has 2734 members. Based on recent traffic it
> >> is really a user / Q&A mailing list, not a place for the
> >> maintainers/steering committees of major projects to speak publicly
> >> with one another (where discussions are public, archived, searchable).
> >> I have observed that there are many discussions happening between the
> >> developers of projects on an ad hoc basis and on ad hoc communication
> >> channels (both private and public). Partly there is no obvious place
> >> for, e.g., the developers of pandas and dask to have a public
> >> discussion, that is not necessarily "partisan" to one of those
> >> projects.
> >>
> >> As another example issue, there is not an obvious place to raise
> >> issues. Back in the day I think numpy-discussion or scipy-user used to
> >> partly serve this purpose, but the centers of gravity have shifted.
> >>
> >> - Wes
> >>
> >>
> >> On Wed, Dec 26, 2018 at 9:41 PM Andy Ray Terrel <andy.terrel at gmail.com>
> wrote:
> >> >
> >> > I'm not completely clear what is being asked for since
> pydata at googlegroups.com already exists. Since NumFOCUS is promoting the
> PyData conference and helping build the brand for the ecosystem, I wonder
> if a home like pydata-dev at numfocus.org would be interesting for folks?
> >> >
> >> > It is also my understanding that there will be a fuller steering
> committee set up for the conferences next year. I propose we do the same
> for the technical structure. As is, I manage the website and github repos
> but there is not much dictating how I manage these.
> >> >
> >> > -- Andy
> >> >
> >> >
> >> > On Wed, Dec 26, 2018 at 6:00 PM Nathaniel Smith <njs at pobox.com>
> wrote:
> >> >>
> >> >> Other examples of discourse used for dev discussion include:
> >> >>
> >> >> - https://internals.rust-lang.org/ -- main dev forum for rust
> >> >> - https://discuss.python.org/ -- potential replacement for
> >> >> python-{committers,dev,users}, still experimental but where a ton of
> >> >> the python governance discussion happened
> >> >>
> >> >> My impression so far is that discourse takes a bit of adjustment
> >> >> compared to mailing lists, but it has a lot of valuable features like
> >> >> multi-quoting, markdown (code blocks, links, ...), solid moderation
> >> >> tools (e.g. if a discussion diverges you can retroactively split
> parts
> >> >> of it out into a new topic), polls (these were incredibly useful for
> >> >> taking the temperature of the community during the governance
> >> >> discussions), ability to reply to messages that were posted before
> you
> >> >> joined the list, configurable notifications (email me everything /
> >> >> email me when a new topic is created / email me a summary weekly /
> >> >> ...), ...
> >> >>
> >> >> -n
> >> >>
> >> >> On Wed, Dec 26, 2018 at 3:41 PM Matthew Rocklin <mrocklin at gmail.com>
> wrote:
> >> >> >
> >> >> >
> >> >> >>> Copying the mailing list
> >> >> >
> >> >> >
> >> >> > Whoops!  E-mail fail on my part.
> >> >> >
> >> >> >>> Discourse is interesting. It seems to be used (at least in
> PyTorch's
> >> >> >>> case) as more of a modern message board for users than a place
> for
> >> >> >>> long-form discussions between project developers.
> >> >> >>>
> >> >> >>> IMHO having a cross-project developer mailing list is probably
> overdue
> >> >> >>> -- I think we can do a better job the next couple of years
> >> >> >>> coordinating (colluding?) with each other. A lot of coordination
> does
> >> >> >>> of course in private, project-level, or other ad-hoc basis. It
> would
> >> >> >>> help to be able to discuss ecosystem-level problems and possible
> >> >> >>> solutions.
> >> >> >
> >> >> >
> >> >> > Entirely agreed.  And I think that an e-mail list is the obvious
> first choice here.
> >> >> >
> >> >> > I'm bringing up discourse as an alternative for consideration.
> This is for a couple reasons:
> >> >> >
> >> >> > I'm slightly concerned that a broad ranging e-mail list that
> encompasses all of PyData would get noisy.  For example I can imagine
> lengthy conversations on visualization or probabalistic programming that,
> while I find important, would likely want to take a pass on.  Having a
> service that includes tags and subscription to those tags may have value.
> >> >> > E-mail list archives tend to collect dust.  If we're doing
> long-range cross-project coordination then those conversations might have
> long term value.  We might want to cross reference them, upvote them,
> subscribe to them, and so on.
> >> >> >
> >> >> > In regards to PyTorch's discuss in particular I agree that it is
> used more as a user forum, which I agree is a different use case than what
> Wes is proposing here.  I mostly pointed to it so that people could get a
> sense of what an active system looks like.
> >> >> >
> >> >> > Regardless, I encourage this conversation to happen with a broader
> set of people.  I believe that other groups are considering these topics as
> well and may have thoughts beyond those that have been expressed here.  I'm
> not sure how best to bootstrap this process, other than an e-mail to maybe
> the NumFOCUS mailing list and perhaps a tweet?
> >> >> >
> >> >> > > There's both a NumFOCUS discourse and zulip, I believe, but
> neither is particularly active. Whether those should be considered possible
> starting points or cautionary tales I leave to y'all.
> >> >> >
> >> >> > Yeah, I should also amend my previous statement from "how about
> discourse?" to "is there anything more appropriate than an e-mail list?".
> Discourse is the service around which I've seen the most activity recently
> but I agree that in our community, it hasn't really taken off that well.
> >> >> >
> >> >> > And just to reiterate, I think that an e-mail list would be
> great.  Just wanted to throw out some other thoughts.
> >> >> >
> >> >> > Best,
> >> >> > -matt
> >> >> >>>
> >> >> >>>
> >> >> >>> > On Wed, Dec 26, 2018 at 8:47 AM Wes McKinney <
> wesmckinn at gmail.com> wrote:
> >> >> >>> >>
> >> >> >>> >> I sent a request to postmaster @ python.o to create
> >> >> >>> >> pydata-dev at python.org. We can also use google groups if
> others prefer
> >> >> >>> >> that
> >> >> >>> >>
> >> >> >>> >> On Tue, Dec 25, 2018 at 3:59 PM Joris Van den Bossche
> >> >> >>> >> <jorisvandenbossche at gmail.com> wrote:
> >> >> >>> >> >
> >> >> >>> >> > Giving the growing ecosysten of data tools (in some way
> related to pandas, but not pandas itself), I am also +1 on such a list. I
> think that would be welcome, and not aware of anything existing.
> >> >> >>> >> >
> >> >> >>> >> > Joris
> >> >> >>> >> >
> >> >> >>> >> > Op di 25 dec. 2018 02:19 schreef Stephan Hoyer <
> shoyer at gmail.com:
> >> >> >>> >> >>
> >> >> >>> >> >> +1 for pydata-dev
> >> >> >>> >> >>
> >> >> >>> >> >> I don't think there's a list quite like this today.
> >> >> >>> >> >>
> >> >> >>> >> >> On Mon, Dec 24, 2018 at 6:11 PM Wes McKinney <
> wesmckinn at gmail.com> wrote:
> >> >> >>> >> >>>
> >> >> >>> >> >>> I'm talking about public archived communication channels
> >> >> >>> >> >>>
> >> >> >>> >> >>> On Mon, Dec 24, 2018, 7:57 PM William Ayd <
> william.ayd at icloud.com wrote:
> >> >> >>> >> >>>>
> >> >> >>> >> >>>> What do you think is missing from the existing PyData
> conferences? I’ve only been to the one in LA but it seemed to be somewhat
> in the direction of what you are asking for.
> >> >> >>> >> >>>>
> >> >> >>> >> >>>> > On Dec 24, 2018, at 3:02 PM, Wes McKinney <
> wesmckinn at gmail.com> wrote:
> >> >> >>> >> >>>> >
> >> >> >>> >> >>>> > hi folks,
> >> >> >>> >> >>>> >
> >> >> >>> >> >>>> > Prompted by some recent discussions I wondered what
> you all think
> >> >> >>> >> >>>> > would be the best venue to have public discussions
> that involve other
> >> >> >>> >> >>>> > open source projects that are generally 1 degree of
> separation away
> >> >> >>> >> >>>> > from pandas. Sort of like "pydata-dev", or something.
> Is there
> >> >> >>> >> >>>> > something like this already that I just missed?
> >> >> >>> >> >>>> >
> >> >> >>> >> >>>> > As context, I'm trying to travel less and go to fewer
> conferences the
> >> >> >>> >> >>>> > next couple of years, and spend more time coding and
> writing, but I
> >> >> >>> >> >>>> > still want to talk with people (asynchronously) about
> things, and
> >> >> >>> >> >>>> > preferably in public.
> >> >> >>> >> >>>> >
> >> >> >>> >> >>>> > - Wes
> >> >> >>> >> >>>> > _______________________________________________
> >> >> >>> >> >>>> > Pandas-dev mailing list
> >> >> >>> >> >>>> > Pandas-dev at python.org
> >> >> >>> >> >>>> > https://mail.python.org/mailman/listinfo/pandas-dev
> >> >> >>> >> >>>>
> >> >> >>> >> >>> _______________________________________________
> >> >> >>> >> >>> Pandas-dev mailing list
> >> >> >>> >> >>> Pandas-dev at python.org
> >> >> >>> >> >>> https://mail.python.org/mailman/listinfo/pandas-dev
> >> >> >>> >> >>
> >> >> >>> >> >> _______________________________________________
> >> >> >>> >> >> Pandas-dev mailing list
> >> >> >>> >> >> Pandas-dev at python.org
> >> >> >>> >> >> https://mail.python.org/mailman/listinfo/pandas-dev
> >> >> >>> >> _______________________________________________
> >> >> >>> >> Pandas-dev mailing list
> >> >> >>> >> Pandas-dev at python.org
> >> >> >>> >> https://mail.python.org/mailman/listinfo/pandas-dev
> >> >> >>> _______________________________________________
> >> >> >>> Pandas-dev mailing list
> >> >> >>> Pandas-dev at python.org
> >> >> >>> https://mail.python.org/mailman/listinfo/pandas-dev
> >> >> >
> >> >> > _______________________________________________
> >> >> > Pandas-dev mailing list
> >> >> > Pandas-dev at python.org
> >> >> > https://mail.python.org/mailman/listinfo/pandas-dev
> >> >>
> >> >>
> >> >>
> >> >> --
> >> >> Nathaniel J. Smith -- https://vorpus.org
> >> >> _______________________________________________
> >> >> Pandas-dev mailing list
> >> >> Pandas-dev at python.org
> >> >> https://mail.python.org/mailman/listinfo/pandas-dev
> >> >
> >> > _______________________________________________
> >> > Pandas-dev mailing list
> >> > Pandas-dev at python.org
> >> > https://mail.python.org/mailman/listinfo/pandas-dev
> _______________________________________________
> Pandas-dev mailing list
> Pandas-dev at python.org
> https://mail.python.org/mailman/listinfo/pandas-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/pandas-dev/attachments/20181229/641ff5de/attachment-0001.html>

More information about the Pandas-dev mailing list