[ruby-core:87425] Re: [ruby-cvs:70678] naruse:r63587 (trunk): Introduce write_timeout to Net::HTTP [Feature #13396]

[ruby-core:87425] Re: [ruby-cvs:70678] naruse:r63587 (trunk): Introduce write_timeout to Net::HTTP [Feature #13396]

[Eric Wong]

naruse@ruby-lang.org wrote:
>   https://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=revision&revision=63587

For OpenSSL, I think you need to expect : wait_readable on
write_nonblock, too.


Anyways, the code for handling partial write_nonblock case is verbose.
One day, I would like to:

1) integrate Timeout into core
2) make all SOCK_STREAM sockets non-blocking by default
3) Make rb_wait_for_single_fd aware of Timeouts

So we can use:

    Timeout.timeout(@write_timeout) { @io.write(strs) }

And no new background threads get spawned.


P.S.: If Ruby were LGPL-2.1+, I would steal the ccan/timer module which
      is optimized for frequently-expiring timers and be done with 1),
      already.

[ruby-core:87509] timeout in core (was: naruse:r63587 (trunk): Introduce write_timeout to Net::HTTP [Feature #13396])

[Eric Wong]

Eric Wong <normalperson@yhbt.net> wrote:
> Anyways, the code for handling partial write_nonblock case is verbose.
> One day, I would like to:
> 
> 1) integrate Timeout into core
> 2) make all SOCK_STREAM sockets non-blocking by default
> 3) Make rb_wait_for_single_fd aware of Timeouts

FYI, I'm close to having a patch ready for 1) and 3);
but maybe 3 is optional, even.

> So we can use:
> 
>     Timeout.timeout(@write_timeout) { @io.write(strs) }
> 
> And no new background threads get spawned.

> P.S.: If Ruby were LGPL-2.1+, I would steal the ccan/timer module which
>       is optimized for frequently-expiring timers and be done with 1),
>       already.

ccan/timer may not be the right tool for the job (more on this
later).