On Thu, 9 Aug 2007, Anlin Zhang wrote:
> Attached is a diff -u out put to show my fix of the tcp connection problem.
> I also insert here since it's not big:
Three problems with this patch:
1. it doesn't follow the current source code style (indent levels)
2. it isn't written to build only for the systems that support these options
3. it contains printf() calls
> The tcp keepalive option I added on the soket will take care of any
> non-responding machine.
First, not all systems support these so I would rather have a more universally
available approach. Possibly in addition to this.
Then, you set the values pretty extreme, imho... I haven't really investigated
this, but is this TCP connection kept alive between the resolves? I kind of
assume so, and then this 2-second keep-alive will create quite some traffic
for lots of applications.
> BTW, how do I directly answer your posting on the web? so that it looks like
> a cascade follow-up?
On the web? This is a mailing list so if you just reply to the mail it'll show
up threaded in all subscribers' mail clients and in the web archives showing
this mailing list.
-- c-ares -- my preferred DNS asynch resolver libraryReceived on 2007-08-11