> Given equal cost path routing like above .
Load splitting routes do not look like this.
BTW I understood your original question only now 8)8)
> How does SIOCDELRT even with the netmask determine
> which of the two above to remove ?
>From the viewpoint of gated (read BSD), the situation where
there are more than one route with one prefix is impossible.
Gated just will be confused by them, but will not die yet.
>From the viewpoint of Linux, it is valid situation,
but only THE FIRST ROUTE is used in routing without
additional hints. It is the first route that deleted
by SIOCDELRT.
> Or does it select the first route in its table ?
Yes. 8)
> Also does SIOCADDRT add routes into it's table in a
> fifo manner or ???
Rather lifo. The last added route will be deleted first.
Actually, SIOC{ADD|DEL}RT is very small subset of the things,
which 2.1 does.
> I think I see the answer to this .
> More specific routes are on -top- (so to speak) and
> the less specific ones are lower . (correct ?)
Yes.
> > > > > does 2.1.x support equal cost path routing ?
Opla! Seems, my mailer is broken. Ocacsinally, I put . at the
beginning and that's result 8)
||||||||||.
> > if someone needs it. Apparently, for my lovely gated-3.6Alpha2 8)
>
> I'd be very interested in a patch against ,
> ftp://ftp.gated.org/net-research/gated/gated-3-5-10.tar.gz
> if available , If not one against 3.6.a2 (a venerable daemon)
> would be good also .
For 3.5 wait until I (or someone) will port netlink to 3.5
I cannot promise, to be honest, I am completely satisfied with 3.6
(especially, taking into account that 3.6 is removed from merit directory
and hence there is no danger, that I will be forced to merge merit
changes 8)8)8))
BTW "venarable" is not quite appropriate word. Seems, 3.5.10 has nothing
new comparing to 3.6 and has not less bugs, than 3.6 had.
Taking into account that 3.6 has netlink patch it is better and newer 8)8)
Alexey
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.rutgers.edu