Daniel Stenberg wrote:
> On Sun, 21 Mar 2004, Joe Halpin wrote:

>
> I approve of the idea and the concept. I have only a few minor nits on the
> patch:
>
> 1. I'd rather see a new command line option than an environment variable for
> the command line tool (perhaps called --tcp-nodelay).
>
> 2. Instead of duplicating 5 lines of code, I'd prefer to have the actual code
> that do the setsockopt() call put into a separate function, and call
> that function from two places (the ipv4 code and the ipv6 code).
>
> 3. errno is not set on windows for network related stuff as one would think,
> use Curl_ourerrno() to get a portable errno. (used if the setsockopt() call
> returns failure)

Ok, I'll make these changes tonight after work. One question though, if
I call Curl_ourerrno() will the return value be useable by strerror() on
Windows as well as Unix?

> Should we include some further details on what TCP_NODELAY actually does in
> the curl_easy_setopt man page?

That's probably a good idea. I'll add more detail to that.

Joe
Received on 2004-03-22