[Xorp-users] Stuff

Pavlin Radoslavov pavlin@icir.org
Tue, 14 Mar 2006 15:26:07 -0800


> I have 3 points.
> 
> 1. The fact that XORP falls over when an interface is actually removed
> (actually stopping a process that creates the tun/tap interface) is a
> biggie for me. Does this not affect anyone else?

Yes, this is not the desired behavior. Please submit a bugzilla
entry about it.

> 2. Is it not possible that, if you try and add protocol pimsm4 int1
> vif int1, and you haven't declared int1 in the interface section that
> it could automatically add that for you? Along with the mfea4, etc,
> etc?

The current model is that (most) entities need to be explicitly
configured, because this is the behavior with least surprises.
Indeed, some leaf nodes have default values, but we want to be very
careful if we want to automatically add more complex configuration
statements. For example, if we automatically add the MFEA
interface/vifs, then someone may ask why not adding the
interface/vif inside the "interfaces {}" section as well.

Once we start going this path, then it becomes more difficult to
preserve the configuration consistency. E.g., if the user deletes
the interface/vif inside the pimsm4 statement, do we automatically
delete the interface/vif inside the mfea4 statement as well?

I am not saying that such optimizations should be forbidden.
Eventually, you should be able to write your own xorpsh/rtrmgr
extentions that would automatically fill-in extra stuff for you, but
we are not there yet, so for the time being we prefer to keep things
simpler.

> 3. I'm trying to set up a multicast network in order to learn more
> about PIM, but it's tricky. I can't seem to find anything about the
> Mbone these days - did that disappear?

Originally, Mbone was about running DVMRP over tunnels.
This has gradually dissolved into running PIM-SM within domains and
MSDP between domains (within the IPv4 context only).
There might be some remaining DVMRP tunnels, but this won't help you
much.

You may want to try M6bone, the IPv6 multicast backbone, though
obviously you would have to run IPv6:

http://www.m6bone.net/

If you are using Linux, then you would have to patch your kernel (or
use Usagi's kernel) to add IPv6 multicast routing support. See the
following URL for details:
http://mailman.icsi.berkeley.edu/pipermail/xorp-users/2005-November/000901.html

Pavlin


> If it has, is there any sort of test network for joining up to?
> If not, would anyone on here like to create one? Grab a BGP AS each, a
> few openvpn/ipsec tunnels around the place, and an IRC channel, and we
> could get one working pretty well, methinks. If you would, please get
> in touch off-list.