cURL / Mailing Lists / curl-library / Single Mail


RE: socks5 and bug item #741841

From: <>
Date: Sun, 15 Feb 2004 18:42:55 -0600

The current code (checking for 0x01) definitely does not work for the
third-party SOCKS5 server that we are using here at work, since it returns

Additionally, I have written similar SOCKS5 networking code that has been
used publically in the client (dnetc) since 1997, I can
confirm that checking for the value 0x05 is what is necessary. We have
never had any complaints that our SOCKS5 client implementation had such an
interoperability problem.

I can confirm that the new cURL behavior of checking for 0x01 no longer
works in my development environment, but patching the check back to 0x05
fixes the problem.

> I moved the discussion around bug #741841 [*] over to this
> list, since it is a nicer place to keep discussions on.
> In this report, Jeff Lawson says the current approach is not
> the correct one. He quotes an RFC section and he seems to be
> right in his claim.
> However: I don't have any socks5 proxy to try on myself. I've
> got a few fixes for this socks5 code in the past from people,
> indicating that they can in fact use the code as it currently
> looks/works.
> Will fixing this as Jeff suggests break the functionality for
> someone else?
> [*] =
Received on 2004-02-16