[concurrency] about the concurrency-sig list

[Jeremy McMillan]

In the olden days, someone on the list would offer to curate a FAQ. The purpose of a FAQ is to squelch debate that is not productive, and I think specifying some guidelines would add value. 

Here's my suggestion. 

1) What sort of discussion is appropriate for concurrency-sig?
a: How to generalize parallel problems (is a discussion topic adequate covered by more general prior discussion)?
b: How to generalize Python approaches/components/solutions to parallel problems (frameworks, multicore, MPP clustering, coroutine/pipeline idiom, event processing, etc.)
c: How to hypothesize performance of an approach to a problem via Amdahl's Law and/or difficulty getting a good implementation
d: How to design and conduct robust testing
e: Analysis of test data
f: Discussion of implications to general knowledge (goto a:)

Discussion should, IMHO, progress through a..f states and then the particular subject can be closed. If you want to know how to get more concurrency in a Python solution to X, that falls under a. If you want to compare X and Y toolkits, that's probably b, but the discussion needs to point to c. Trouble with any given system are under c and d. Bring data everyone loves e! With new data, we can cut loose a little bit and prune our collective dogma together :)

2) What sort of discussion is inappropriate for concurrency-sig?

If it doesn't fit one of the states in 1), or it prevents discussion from progressing through those states, we're not interested.

On Nov 15, 2011, at 5:00 AM, concurrency-sig-request at python.org wrote:

> From: David Beazley <dave at dabeaz.com>
> To: py concurrency sig <concurrency-sig at python.org>
> Subject: Re: [concurrency] about the concurrency-sig list
> Message-ID: <49492659-F1CE-4D84-BBF1-143365BFCADD at dabeaz.com>
> Content-Type: text/plain; charset=us-ascii
> 
> Pete's wiseguy thread comment not withstanding, I'll freely admit that I'm usually not interested in participating in online discussions about concurrency due to excessive religious dogma backed by a lack of empirical data or experiments.   IMHO, there needs to be more level-minded experimentation, analysis, and discussion.  If this is what this list might be about, then it might be worthwhile.
> 
> Cheers,
> Dave
> 
> 
> On Nov 11, 2011, at 2:54 PM, Peter Portante wrote:
> 
>> Perhaps we would want to avoid dogma, and have this list provide understanding of how things work and pros & cons?
>> 
>> On Fri, Nov 11, 2011 at 3:47 PM, Jesse Noller <jnoller at gmail.com> wrote:
>> 
>> 
>> On Friday, November 11, 2011 at 3:44 PM, Peter Fein wrote:
>> 
>>> Hi-
>>> 
>>> I started this list & the concurrency section on the wiki
>>> http://wiki.python.org/moin/Concurrency after taking one of Dave
>>> Beazley's classes. Other folks & I wanted a place to discuss the kind
>>> of issues Brad discusses below, in a general and cross-toolkit
>>> context. I don't know why it never really took off, though I'd love to
>>> see it have more life.
>>> 
>>> Other possible topic includes: Python under Hadoop, MPI, data
>>> processing pipelines (generators).
>>> 
>>> Should we just establish "Threads: you're doing it wrong" as a ground
>>> rule and be done with it? ;-P
>>> 
>>> --Pete
>>> 
>> No, given that "Threads: you're doing it wrong" is patently incorrect except for certain cases. I use them more than I use multiprocessing. :)
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/concurrency-sig/attachments/20111115/e3331776/attachment.html>