16 jun 2010 kl. 08.48 skrev Andrei Pelinescu-Onciul: > On Jun 16, 2010 at 08:41, Olle E. Johansson <[email protected]> wrote: >> >> 15 jun 2010 kl. 16.45 skrev Andrei Pelinescu-Onciul: >>> raw sockets: udp send will use now raw sockets if enabled >>> >>> A raw socket, that will be used for sending, is now initialized on >>> startup. The operation can succeed only if sr is started as root >>> or with CAP_NET_RAW. >>> If the initialization fails and the raw sockets are forced on from >>> the config (udp4_raw = 1), ser won't start. If the raw socket are >>> in auto mode (udp4_raw = -1), sr will start, but with the raw >>> socket send part disabled (normal udp send will be used). >>> >>> If the raw socket part is not disabled at startup, it can be >>> turned anytime on/off via the runtime cfg. framework. >>> >>> Note: the raw socket send is supported only over ipv4 and for now >>> only on linux. >> >> Andrei, >> Can you explain the benefit of the raw sockets too? This sounds very >> interesting, but I don't see the reason why you want this. I am very >> curious! > > It's experimental stuff for now. > See http://lists.sip-router.org/pipermail/sr-dev/2010-June/007760.html > for a little more info. More details after I get to do some performance > testing. > > Also parts of the code will be used in the future for icmp based > blacklisting and maybe for some kind of udp only transparent proxy or load > balancer modules (icmp much more likely to happen though). > Ahh. So we will react to ICMP unreachable messages. That's a huge improvement!
Thanks. /O _______________________________________________ sr-dev mailing list [email protected] http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev
