This is to announce the release of Pound v2.0.5.

Changes in this version:

- added support for the Hoard library. If the Hoard library is installed
Pound will use it instead of the normal memory allocation functions
found in libc. Hoard was specifically designed to support multi-threaded
applications, and it seems to do a good job of it.

In our testing we saw no problems, but Hoard may not be completely
compatible with the libc malloc - so we would really appreciate
additional testing.

The software is at version 2.0.5 (beta quality). Further testing 
(especially under heavy loads), improvements and suggestions are
welcome.

If no problems surface we'll freeze this as version 2.1 by the end of
next week.[...]

This is to announce the release of Pound v2.0.6.

Changes in this version:

- added support for back-ends that work over Unix-domain sockets, as per
Cliff Wells' suggestion. This feature needs quite a bit of testing - I
hope we got it right.

The way it works is that if the back-end address cannot be resolved (as
an Internet address) Pound will try it as a path to a Unix-domain
socket. See the man page and source for details.

- integrated Russel Valentine's patch for updating the time of last
access for sessions. This is a welcome bug-fix and should solve problems
for people who had been "losing" sessions.

The software is at version 2.0.6 (beta quality). Further testing 
(especially under heavy loads), improvements and suggestions are
welcome.[...]

On Sat, 2006-05-20 at 18:35 +0200, Robert Segall wrote:[...]

Well, I knew Pound was fast, but I had no idea its developer was that
fast, otherwise I would have just waited before implementing my own.
Oh well, I can at least compare the two and see if I missed anything :P

Seriously though, thanks a lot for the fast turnaround on this, this
will make my life much simpler.  I'll let you know how the testing goes.

Regards,
Cliff

On Sat, 2006-05-20 at 18:35 +0200, Robert Segall wrote:[...]

Almost =)

config.c, line 366:

366c366
<             if(res->domain != PF_INET && !has_port) {
---[...]


You implementation shows the same issue with Lighttpd's mod_cgi as mine
does, so that's a nice outside confirmation of the issue with lighty.

Will continue testing.

Regards,
Cliff

On Sat, 2006-05-20 at 18:35 +0200, Robert Segall wrote:[...]

In case you are interested, I have a patch that fixes the logging in
http.c so that unix domain sockets don't show up as 0.0.0.0, but rather
the path to the socket.  I'll have to change it slightly to work with
your implementation, but I'll pass it along later tonight.

Regards,
Cliff