Author gvanrossum
Recipients christian.heimes, felipecruz, giampaolo.rodola, gvanrossum, meador.inge, neologix, pitrou, rosslagerwall, sbt, vstinner
Date 2013-08-21.16:15:14
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <CAP7+vJKP=mkOhqS6VnbLf3bjRnM8RdAVYCbUDdkgKznM_rMyWQ@mail.gmail.com>
In-reply-to <CAMpsgwbdZTFE2bGzSX5HWtDomxaxDWKxkPakbDVBxGx_aF2=-A@mail.gmail.com>
Content
On Wed, Aug 21, 2013 at 2:37 AM, STINNER Victor <report@bugs.python.org>wrote:

>
> STINNER Victor added the comment:
>
> I like the idea of renaming select to _select, and add Lib/select.py
> for the high-level wrapper (Selector). In my opinion, adding a module
> just for 5 classes is overkill. Selector classes are a very thin
> abstraction over the low-level objects. A new module can be justified
> for something more evolved, something like an event loop with
> callbacks and a scheduler, something like the PEP 3156 (tulip) :-)
>

OK, I'm just going to impose my decision. I want a separate selectors.py.
The rename business feels unpleasant, *and* the interface abstraction level
is different.

> This issue is for code written in Python. Is it a similar issue for
> code written in C? internal_select_ex() of Modules/socketmodule.c and
> check_socket_and_wait_for_timeout() of Modules/_ssl.c uses poll() and
> select().
>

Let's leave those alone. AFAIK they are mostly to simulate timeouts in
*blocking* code.
History
Date User Action Args
2013-08-21 16:15:14gvanrossumsetrecipients: + gvanrossum, pitrou, vstinner, giampaolo.rodola, christian.heimes, meador.inge, neologix, rosslagerwall, sbt, felipecruz
2013-08-21 16:15:14gvanrossumlinkissue16853 messages
2013-08-21 16:15:14gvanrossumcreate