On Thu, 28 Aug 2003 22:37:06 +0200 Henrik Nordstrom <[EMAIL PROTECTED]> wrote:
On Thursday 28 August 2003 17.37, Evgeny Kotsuba wrote:
My squid is not uzing signals for IPC, instead it uses named pipes so now I add option to command line parameters to send to working squid (section, level) as in debug_options in config.
May be this idea will be interesting to mainstream squid developers.
It is. The signal approach is getting really limiting. There is many actions we want to be able to perform on a running Squid, but nobody has been motivated enough to write a pipe control channel.
If you have written one, please submit it for inclusion yesterday ;-)
I have a old patch (from the 2.2.PRE1 timeframe) to implement selective acl reloading which is a immediate candidate ot use this named pipe. The original patch was rejected as it relied on cachemgr to trigger the action and we do not want active functions in cachemgr..
I have wrote already about it - my port of 2.5 to OS/2 is a bit different from mainstream - I use OS/2 pipes/API as well as a number of differences (C++ compiler, limited use of classes and threads, color ANSI text out for non-daemon mode text out on errors/warnings, sound beeps on error/warnings etc.)
SY, Evgeny Kotsuba
