Of course you're right, but let me play devils advocate for a bit:
I am not sure how important it is to separate fast-hello from route
announcements for hosts. We may have <= 10 addresses on a host. Where would
be have a problem ? processing in the receiving router ?
Something as simple as RIP also makes it very easy to embed it as
a library into apps, which is particularily good if you have an (anycast)
application specific address and you actually want to make sure the
address is only announced when the app is alive.
BFD takes modularily even one step further and if we used that, then
the benefit of hello/route advertisement separation in the routing
protocols themselves becomes mood. But it adds even more code and
complexity.
Fast Hellos are good. Actually i'd love to have something thats even
better and can also double to recognize lousy links. Oops: marginal links.
But then there is all this talk about low power wifi clients as well and
that makes active probing supposedly prohibitive.
IMHO, the most important question is:
Whats the option that could likely be proliferated into eg:
MacOS/Windows/Android ? Lets just say with <= 5 sec failover time.
Cheers
toerless
On Sat, Feb 21, 2015 at 07:52:44PM +0100, Juliusz Chroboczek wrote:
> > I think i had tuned down the RIP hello interval.
>
> Impossible. It was the Update interval that you had tuned down.
>
> > I probably would prefer not to use one of the real routing protocols, but
> > something lightweight. RIP is a stupid routing protocol but just to
> > announce aliveness of host interfaces it is pretty ook.
>
> RIP(ng) has a major flaw: it doesn't have explicit Hellos. Reachability
> is established by receiving Update packets, which have a size that is
> proportional to the number of advertised prefixes.
>
> Modern routing protocols (at least IS-IS, OSPF, EIGRP and Babel) have an
> explicit reachability sub-protocol, that consists of explicit Hello
> exchanges. This allows sending Hellos much more often than you send
> Updates or topology information. For example, in Babel a packet
> containing just a single, unextended Hello has a payload of just 12 bytes,
> independent of the size of the network.
>
> (In principle, you could hack a Hello protocol into RIP by interpreting
> empty RIP packets as reachability indications. But that would no longer
> be RIP, right?)
>
> > Of course something defined specific for this purpose (host address
> > aliveness or the like) would be most clean IMHO.
>
> I believe that BFD (RFCs 5880 and 5881) is the IETF Standard Track
> reachability detection protocol.
>
> -- Juliusz
--
---
Toerless Eckert, [email protected]
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet
