Re: no_chache directive

On Fri, 13 Nov 1998, Alex Rousskov wrote:

]> So, we are back at recursing through the subdirectories ourselves, and
]> browsing the head of each file, I guess?
]
]Or we need a small patch to storeKeyPublic() to make it return a
]encoded-in-the-URL entry KEY if the url is in a well-known format. For
]example, when the url is
] /squid-internal-key/<PUBLIC_KEY>
]storeKeyPublic() returns
] <PUBLIC_KEY> after ascii->bin conversion.
]
]Then we can use keys in "normal" requests to Squid. There may be one or two
]other places that will have to be modified, but nothing major, I guess. Will
]that be a useful patch / compile-time option?

Let me see, if I understand you. You suggest to extend squid so it can
kind of reverse lookup the key (e.g. 0000BC2171E4AAE62C00D46245F5F36E) and
return the URL associated with it. Then we could just open a socket A, ask
for the mgr:object URL. With each key from socket A, we could use a
persistent connection socket B, to ask for the associated URL. We could
also use the persistent B to send the PURGE request, or use, for design
purposes, a separate persistent connection C.

So, yes, such an extension of a 'reverse lookup' of squid keys would be a
feasible patch, if that was you intend. I don't believe it ought to be
just a compile-time extension.

Le deagh dhùrachd,
Dipl.-Ing. Jens-S. Vöckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726
Received on Mon Nov 16 1998 - 02:59:07 MST