Re: newhttp branch

From: Robert Collins <robert.collins@dont-contact.us>
Date: Fri, 27 Apr 2001 22:24:53 +1000

----- Original Message -----
From: "Adrian Chadd" <adrian@creative.net.au>
To: "Robert Collins" <robert.collins@itdomain.com.au>
Cc: <squid-dev@squid-cache.org>
Sent: Friday, April 27, 2001 10:00 PM
Subject: Re: newhttp branch

> > objects. Header parsing is about 80% done - waiting on the
(storemanager
> > to handle headers outside of the data stream) || (the broker
interface).
> >
> > So I guess I'm really saying that between us, if you can add the
> > appropriate logic to modio, or to content_processing if I move
> > rbcollins.filters to content processing (where it should be) and
then
> > merge modio down there as well; the only significant code mangling
> > needed should be in the store* interfaces. - do you want to
collaborate
> > on this bit?
>
> Ah. See, the reason I'm doing the newhttp stuff is to write a set
> of http routines that don't rely on the current behaviour in the
> storage manager and network IO code.

I think I see. The goal is to get a clean interface and then bring in
the eventloops and modio stuff?

> I mean, personally, we shouldn't _need_ that much code to implement
> a basic HTTP proxy, assuming I can recycle the Http*.c routines.
>
> So, how about this:
>
> * I work on bringing the http client/server code into using the
> broker interface so it implements a HTTP/1.1 proxy, with the
> headers seperate from the body (this implies the "broker")

Sure.

> * once this is working we look to bring in the modules framework.

The module framework is really kinda irrelevant (It abstracts the
dynamicness needed to add in the clientlets and register protocols etc -
so it's not a huge issue either way).

Its the filter code itself - the clientlets - that was the difficult
thing to get ironed out. The reason I've been drilling into this is to
try and minise the rework needed to leverage that time-investment :].

> * once the modules framework is in place we look at turning our
> codebase into a cache once again.

As in get the broker to store the objects? Ok.

> I'll go and look at the branch diffs now.

I'll bring rbcollins over to content processing this weekend. Promise :]

Rob

>
>
>
> Adrian
>
Received on Fri Apr 27 2001 - 06:24:30 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:50 MST