Re: Forward architecture.

Adrian Chadd wrote:
> Thats basically what Robert was proposing with his method changes
> in squid-3 a few years ago.
>
> Of course, what you really want to do is break out all of the magic
> url manipulation stuff that happens for acceleration, because -that- stuff
> gets rewritten far before the forwarding manager gets its hands dirty..
>

Code references?

Amos

>
> On Sun, Feb 03, 2008, Amos Jeffries wrote:
>> Well, kick me down if this is a brainfart instead of a good idea.
>>
>> While checking and re-working the use of HttpRequestMethod it occured to
>> me that anything startign with a pure ascii word would now be identified
>> as a valid METHOD_UNKNOWN. (nee. HttpRequest::sanityCheckStartLine)
>>
>> Also, that what I have seen of during debugging attempts at
>> transparently passing P2P and IM traffic. That largely those protocols
>> start with a plain-ascii method not belonging to HTTP.
>>
>> The idea that comes to me immediately is to make the forward handle a
>> dispatcher/manager where the currently compiled additional-modules
>> register themselves against various parameters (method, url-protocol,
>> etc) and the forwarder hands of the connection data to them after
>> checking these details. Defaulting to HTTP when unknown.
>>
>> This would enable easily plugged in extras to squid-3.
>>
>> Feedback?
>>
>> Amos
>> --
>> Please use Squid 2.6STABLE17+ or 3.0STABLE1+
>> There are serious security advisories out on all earlier releases.
>

-- 
Please use Squid 2.6STABLE17+ or 3.0STABLE1+
There are serious security advisories out on all earlier releases.