Re: Concurrency, again

[Ted Zlatanov]

On Tue, 25 Oct 2016 19:04:42 +0200 Stefan Huchler <address@hidden> wrote: 

SH> Are you willing to invest much time in implementing your solution? It seems
SH> to be similar to the commercial world, where people say "put your money
SH> where your mouth is", just in this case "put your code where your mouth
SH> is".

SH> I think people would be more convinced if there is some commitment,
SH> cause else you can formulate the nicest thing and nothing will happen.

Commitment and personal investment are not enough to make a code change
convincing.

Please don't forget that writing code is a small (10-15%) part of the
overall cost. Documentation, backwards portability, handling bugs
(including bugs caused by poor documentation or poor API design!),
technical debt, and ongoing maintenance are costs that must be
considered by the Emacs maintainers for any proposed change. The scope,
therefore, matters: the wider the scope of the change, the more risk and
cost it may incur.

In that light, I think Philipp's proposed libtask wrapper may be a good
direction because the risk is low (it relies on existing well-known
libraries) and the cost is low (it doesn't change internals
significantly). Looking at the concurrency-libtask branch, the API
design of the channel piece, at least, seems reasonable: it's well-known
to Go programmers and provides a nice message bus facility.

I don't know how Philipp's proposal interacts with the other concurrency
proposals, though.

HTH
Ted