On Sep 3, 2014, at 4:44 PM, Michael Richardson <m...@sandelman.ca> wrote:
> My logic is that if we have the switch/env-var to force libpcap to use a
> particular method, that permits not only benchmarking, but also isolation of
> bugs easier.
OK, I'd vote for an environment variable, so as not to
1) add a new API that would implement platform-specific behavior (which
would encourage people to use it, rather than getting the underlying problems
fixed);
2) require that programs *using* libpcap add code to use that API and
provide UI to specify platform-dependent options to control the capture
mechanism.
Hopefully the environment variable will also be annoying enough to use that
people will report problems such as the TPACKET_V3 problems and, if they
require OS fixes, put pressure on developers to get them fixed, rather than
having tweaking capture options such as that being "the new normal".
_______________________________________________
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers
