Hey,
On Fri, 9 Aug 2002, Adrian Chadd wrote:
> Ah. yes. Well, if you wanna do html filtering, you'll need to
> code your filter in such a way that if you discover a partial match
> at the end of a buffer, to keep the buffer around somewhere until
> the next buffer comes along so you can do a 'complete match'.
What we tried is a count so it adds up the matches after each chunk. Once
the count of "keywords" is satisfied, it spits a meta refresh to the
client store. Problem is the client gets "some" of the request, then boom
they get a refresh to an ACL blocked URL. This is such a bad hack.
It most likely will leak memory now, because we're no longer
storing/cleaning up the data read in from the web server.
> You _COULD_ also, with some of the stuff that'll go into squid-HEAD
> soonish, write a filter module that buffers all text/html content
> until the complete amount has been read in, and _THEN_ send it to
> the store/client, without affecting other stuff.
This sounds like the better way to do it. Are you guys planning on
something similar, or should I go ahead and just do it??
-- Joe
>
>
>
>
>
> Adrian
>
Received on Fri Aug 09 2002 - 08:05:47 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:16:02 MST