On Mon, Jan 23, 2012 at 9:36 AM, Vitalii Tymchyshyn <tivv00@gmail.com> wrote:
> 23.01.12 16:02, Dave Cramer написав(ла):
>
>> More recently a patch has been presented with the subject NIO support.
>> This patch in fact only provides an alternative implementation of
>> setTimeout. The author has promised an NIO implementation later.
>
> Small correction for everyone to understand it right: the patch does contain
> NIO support for V3 non-secure connections. V2 and secure requires extra work
> I am going to perform in the nearest time. I posted to the list early
> because I simply wanted to get feedback as early as possible.
>
> Note that NIO itself do not provide any specific benefits (may be
> performance?). What you get is much better control on IO, comparing to using
> streaming. So, you can get timeout processing in same thread, you can see
> when your buffers overflow and react accordingly, you can see when your SSL
> data is available for reading.
>
> As far as I can see, introducing NIO do not require global changes. The
> tricky thing is handling read during write, but this is additional feature
> NIO makes possible.
It is however making changes at the very core of the code. I'm not
saying it isn't good or bad, just that it will require considerable
testing.
Dave Cramer
dave.cramer(at)credativ(dot)ca
http://www.credativ.ca