[j-nsp] Independent /32s for Interfaces - anybody doing that?

[Chris Kawchuk]

Just to add to this....

On 31 May 2016, at 9:31 pm, Vincent Bernat <bernat at luffy.cx> wrote:

> Unfortunately, the support vary widely accross vendors. I believe the
> support is pretty good with Cisco. With Juniper, it really depends on
> the equipment. The MX has pretty good support, but has some limitations
> (for example, static BFD + unnumbered doesn't work). 


Doing ip-unnumbered/OSPF on interfaces and also enabling LDP in the interface -- every now and then the LDP doesn't survive reboot, and have to "kick" (clear ldp neighbors) the LDP sessions for inet.3 to correctly populate on the box once OSPF is back up and established. It's once of the reasons I have never widely deployed using ip-unnumbered w/OSPF and MPLS in anything other than my lab environments. Happened on 13.x through 15.x on MX80 and MX480s with various RE's and SCBs/SCBEs. Never got around to opening a PR on this though...

Agreed is great for lab and quick prototyping so I don't have to fiddle around with assigning /30s and /31s when Im trying to do something higher-layer (L3VPN/VPLS/eVPN/iBGP/etc..)

Juniper produced a J-Learn called "MPLS Plug and Play for Metro Ethernet" (google for it) from many years ago (2007) which explicitly showed how to setup your devices to automatically create an IGP/MPLS/LDP mesh in a metro (for L2CKTs)...although it works for any topology of your choice.  Heavily uses the ip-unnubered and the lo0.0 inheritance schemes discussed in this thread; so should be widely supported if they were pushing this....

- CK.