On Mon, 8 Apr 2002 Gereon.Steffens_at_OnVista.de wrote:

> From my current need, what I'm looking for is a API that allows me to do
> something like this:
>
> 1) prepare some sort of container (array/stack/queue/list, whatever) for N
> URIs that I want pipelined
> 2) for (i=0; i<N; ++i) { prepare request_container[i] }
> 3) call the equivalent of curl_easy_perform on the container, which
> returns a N element container of responses
> 4) for (i=0; i<N; ++i) { deal with response_container[i] }
> 5) destroy both containers
>
> In our specific application, N is actually pretty small - more than 6 or 8
> pipelined requests don't really improve performance.

I don't see why this scenario *requires* pipelining. Sure, using pipelining
is a performance boost when doing multiple requests to the same server, but
you can just as well perform this without pipelining until libcurl gets the
proper code for it.

> Is there a (vague) timeframe when you plan to add this? I'd hate to have to
> learn libwww :-)

I've given up trying to estimate dates for future features (as I've failed
too many times). It will happen sooner the more people that help out bringing
it. I try to steer the curl boat, I can't shuffle all the coal too.

If you need pipelining really soon, you either go elsewhere and look for it,
or you join in and start contribute! ;-)

-- 
    Daniel Stenberg -- curl groks URLs -- http://curl.haxx.se/