On Sat, Sep 22, 2007 at 10:59:29PM +0200, Daniel Stenberg wrote:
> I checked around, and at least the FD_CLOEXEC usage needs an #ifdef
> FD_CLOEXEC. People seem to use a different approach for Windows, seen
> partly in a patch such as this (for gcc):
>
> http://gcc.gnu.org/ml/gcc-patches/2005-03/msg02418.html
OK. It should be relatively easy to add, though.
> Also: is disabling the Nagle algorithm unconditionally really always a good
> idea?
It's a good question. I've tried digging through the internal logs here, but
I can't really figure out why it was first added. How often would you want to
batch DNS requests in the first place, though?
> And perhaps more importantly, it would probably be nice with a little
> comment next to the Nagle disabling call that explains why that code is not
> in the configure_socket() function (I assume that is because Nagle's for
> TCP only, while the configure_socket() function works for UDP socket as
> well?)
Yes, that sounds reasonable. (I was wondering why for a while, but forgot to
look it up before submitting the code :-) )
> We'll enjoy getting these changes. Even if some of them change API or
> whatever, just bring them on and we'll discuss how to deal with them on a
> case by case basis!
OK, here goes the first one (next message). :-)
/* Steinar */
-- Software Engineer, Google NorwayReceived on 2007-09-25