Chuck Murcko wrote:

> So, who's here? Tony?

And me - lurking still.

> My thoughts: I'll get the old code working nocache for HTTP 1.1 and 
> CONNECT, and we'll see what happens there about sticking that bit 
> back into the beta, for now.

I firmly believe that the proxy function and the caching function should
be completely separated from each other. The proxy is a pain in the butt
because the whole caching story makes it so complicated.

Making the proxy a dead simple pure pass-through proxy should be step
one, which can be incorporated into the beta again. Being simple it
shouldn't be too complex to maintain, and should make the
every-six-months-proxy-must-die threads go away.

> Anyone who wants to work on the general purpose http-2.0 caching?

I have a proposed design for a generic caching module for Apache.

This design has nothing whatsoever to do with the proxy code, which
becomes simply "yet another provider of web content".

The design is also firmly based on HTTP/1.1 - The HTTP/1.1 protocol
describes in detail how a "public cache" should work, covering all sorts
of details about refreshing objects in the cache, etc. There is no point
in reinvented some other caching scheme - HTTP/1.1 will work well.

The caching design also covers all other content - server side includes,
mod_perl, mod_cgi, who cares - the cache just sees it as 
yet-another-data-stream-that-I-either-cache-or-dont-cache.

> Thoughts? Or just silence, after last week's invigorating threads about new 
> proxy design?

I would be really keen to see some comments on the design. It's been
posted before, but I'm not sure if people looked at it in quite as much
detail. Deciding on a basic framework at the outset will make building
the proxy and cache modules much easier.

Of course the catch is that the design doc is currently sitting on my
currently inaccessible Linux machine - when I get the thing restored
tomorrow I will post it here with some explanatory comments.

Regards,
Graham
--

Reply via email to