Re: very small packages - merge into general/miscelleneous or move into core

[Olaf Till]

On Mon, Jan 27, 2014 at 01:32:59PM +0100, c. wrote:
> 
> On 27 Jan 2014, at 11:03, Olaf Till <address@hidden> wrote:
> 
> > since "parallel" seems
> > notorious to be problematic to install on non-standard operating
> > systems.
> 
> What OS are exactly are being problematic with "parallel"?
> 
> If you refer to windows systems

I do.

> that it should 
> not be a problem to join "parcellfun" into parallel
> because AFAIK the use of "parcellfun" is not recomended on 
> windows either because "fork" is not very efficient there.

It may be inefficient, but the processes need to be created only once,
so for time-consuming tasks parcellfun might still pay off even on
windows. "parallel" currently doesn't compile on mingw, and even if
this is mended by adapting to winsocks, new problems could occur once
I start introducing TLS in addition to bare sockets.

So I'd still hesitate to move these functions from "general" to
"parallel", at least if I have to decide on that on my own.

OTOH the functions naturally belong rather into parallel than into an
extra package. The idea to add you as an additional maintainer there
seems good to me, too, since you seem to care about parcellfun and I'm
not too eager to have additional responsibilities. (I hope the
situation would be better for you than in "general", although
"parcellfun" would again be in the same package with "foreign"
functions.)

It's a difficult decision. And what about the original discussion of
(chiefly) you and Carnë, is there already some consensus, or is the
current problem sufficiently unrelated to treat it separately?

Olaf

-- 
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc
Description: Digital signature