cURL / Mailing Lists / curl-library / Single Mail


Re: connection pooling?

From: Daniel Stenberg <>
Date: Tue, 8 Jan 2002 16:30:01 +0100 (MET)

On Tue, 8 Jan 2002, rick vaillancourt wrote:

> I have a C++ service that wants to make a url call to get some data from
> another service...
> I figured rather than incuring the hit of init every time, I could go at a
> cached curl_handle?

That could make sense. No matter if you pool handles or not, curl is always
best used when not doing cleanup/init but instead just re-using the same
handle over and over again.

> I was thinking of creating some kind of factory class that had like 10
> init'd curl_handles and a state flag available/notavailable.
> C++ service asks factory for first available curl_handle and makes the
> call.

There might be reasons why you need to consider how you go ahead and
implement this:

Each "easy-handle" keeps the last N connections open and alive if possible
for persistant connection reasons. Thus, you probably want to re-use the very
same handle again when re-issuing a request to a host a second time...

Doing this with many requests for each handle may also result in quite a
large number of connections kept alive, which isn't a problem as long as
you're aware of it.

> I wanted to check if someone had something like that out there already.

Not that I know of.

    Daniel Stenberg -- curl groks URLs --
Received on 2002-01-08