Dan McDonald wrote:
On Sat, Jun 27, 2009 at 12:23:19AM -0700, Ben Rockwood wrote:
<SNIP!>
So my question is... given that it seems switch vendors such as Force10
aren't supporting IRDP any more, and RIPv2 is generally considered taboo by
network administrators, what options do I have for router discovery?
I'm pretty sure the latest version of in.routed performs all of the tasks
in.rdiscd used to do...
Right, it does ICMP Router Discovery automatically and will
switch if RIP-2 is present.
Of course, OpenSolaris also comes with IS-IS, OSPF, and BGP
(from Quagga), so if you want something a little more modern,
that's there as well. And if you're using DHCP, that can
establish a default route.
I guess the real question is: if you wanted router discovery,
but you don't want ICMP Router Discovery or RIP-2, then what
exactly do you want?
For what it's worth, I'd recommend RIP-2 over router discovery.
The complexity is essentially the same, and the capability and
reliability is higher. I think the fear of RIP comes from a
number of factors:
- A certain large router vendor has always had a less than
perfect implementation, and thus has told customers that
the protocol itself is flawed.
- Some people think that if you're router and you speak RIP,
then you're "insecure." Of course, any competent
implementation (and most others as well) can set RIP up to
be broadcast-only, so that all received routes are ignored.
And it has exactly the same authentication mechanism as
OSPF.
- Others erroneously think that its timers somehow cause
latency or oscillation in routes. It'd take a longer message
to refute that one, but it's basically bunkum. Decent
implementations produce triggered updates and don't just
batch up events.
- Because RIP-1 was moved to "Historic," some think that the
entire protocol is unusable.
In fact, RIP has a number of advantages over other protocols,
so it still has a place:
1. Unlike OSPF, the sender needn't keep track of receivers.
OSPF requires that each DR has storage proportional to the
number of listening nodes, so that it can retransmit when
necessary. RIP doesn't require that at all. The receivers
call all be passive. It scales nicely for large numbers of
host-type machines.
2. Unlike ICMP Router Discovery, it gives you actual network
routes. You can still emulate Router Discovery by using
just a single default route with RIP -- in other words, it's
a perfect superset of Router Discovery, and thus RD really
has no design purpose whatsoever. (The claimed benefit is
that it's simpler. If you look at the two protocols even a
bit, that's obviously nonsense. If you're a really lazy RIP
receiver, you can just treat all senders as being default
routers and ignore the route contents! Our BSD-based in.routed
can do that as well, if you want.)
So, as far as I can see, it really comes down to whether you want
to design your network based on emotion or technical evaluation.
_______________________________________________
networking-discuss mailing list
[email protected]
