curl-library
Re: CURLOPT_READFUNCTION performance issue
Date: Mon, 29 Jul 2013 13:53:10 +0200
>>> it will not benefit a carefully written application that takes proper care of buffering; the algorithm has either no effect, or negative effect on the application.
>>
>> Does curl take "proper care of buffering"? If yes, could it be beneficial to set CURLOPT_TCP_NODELAY per default?
>
> It can only do so much. You could argue that in this particular case of using the callback method to POST a mere 10 bytes is a rather inefficient approach to buffering...
>
> I'd rather not change the default just based on some wild speculations though, so if someone truly wants to argue for altering the Nagle on/off default switch somehow, I'd ask for data and statistics to back it up first!
I did some tests, but I could not back up the hypothesis, that disabling the Nagle algorithm improves the performance in FTP and HTTP.
Most of the time it even made it a little worse..
I tried to do some tests with tcptrace too but benchmarking this correctly is really hard..
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2013-07-29