This is similar to a discussion we had on jackit-devel a long time ago about whether to have "internal" or "external" clients to the jack infrastructure. We found that the internal clients were more efficient because of problems with shm and context switching between processes, but the icap mode is attractive because it has a standard for an API, or at least proposed standard, and vendors can implement icap and we will win with squid. Although the streams mode would be preferred, and undoubtedly it would be faster, it would require special code to hook into squid, no?
_J
>>> Henrik Nordstrom <henrik@henriknordstrom.net> 11/20/06 6:32 PM >>>
tis 2006-11-21 klockan 00:30 +0100 skrev Henrik Nordstrom:
> mån 2006-11-20 klockan 07:58 -0500 skrev Jeremy Hall:
>
> > I can say from personal experience using the filters patch that it needs work and I would like to see option 2, the icap mode, further developed.
>
> Further agreed, and is why I listed Squid-3 with ICAP as the first
> alternative.
Oh, I didn't. Squid-3 client streams was the first.. oh well.
1 or 2 is good. 1 for inline processing in the same process, 2 for
out-of-band processing in a separate process (or even server..).
Regards
Henrik
Received on Tue Nov 21 2006 - 06:21:22 MST
This archive was generated by hypermail pre-2.1.9 : Wed Nov 29 2006 - 12:00:05 MST