Re: client_side and comm_close

From: Alex Rousskov <rousskov@dont-contact.us>
Date: Sun, 20 Apr 2008 10:44:57 -0600

On Sun, 2008-04-20 at 16:51 +0200, Henrik Nordstrom wrote:
> sön 2008-04-20 klockan 00:38 -0600 skrev Alex Rousskov:
>
> > This thread is (was) dedicated to the discussion whether _adjusting_
> > v3.1 roadmap to add client-side cleanup is a good idea.
>
> And I say no on that. The current code works kind of, and the small
> problems we have can be patched up for 3.1.STABLE.

Noted. Since nobody advocated and volunteered for doing the client-side
cleanup now, I think we should plan doing it for v3.2. If, based on Amos
4-step plan, we decide there may be time to do this in v3.1 after step
#2 is completed, we can always revisit this question. If nobody commits
to do this in v3.2, the feature will be pushed back to v3.3, as usual.

I have added a Feature page to place this work on Squid3 roadmap:
http://wiki.squid-cache.org/Features/ClientSideCleanup

> What is needed for the server component (client_side*) is a major
> cleanup of the server code + store + how it interacts with forwarding.
> It's a faily large change and beyond what we should aim for in 3.1. 3.2
> is more likely, maybe 3.3. Quite likely it needs a complete rebuild of
> the client_side* code (and rename to http_server), only reusing the
> parts that makes sense. It's not really that difficult job that code
> needs to do, but the way it's currently being done makes a big mess of
> it.
>
> client_side interactions with comm_close is just a tiny tiny part of
> that, and looking into that before we have defined how comm_close really
> should behave isn't meaningful imho.

I agree with the technical comments. As for the schedule, I hope we do
not have to wait for v3.3, but it is too early to say.

Thank you,

Alex.
Received on Tue Apr 22 2008 - 13:12:26 MDT

This archive was generated by hypermail 2.2.0 : Wed Apr 30 2008 - 12:00:07 MDT