Add-path support

[netixx]

Hello,

I am trying to implement add-path capability negotiation implementing caps.Cap interface.

However I realized that more core work need to be done do support add-path, because the Reach and Unreach need to include a path ID attributes.

Do you have any plan already to support add-path ?

[pforemski]

Hello, and thanks for your interest in this project!

Yes, add-path would require more work for the reasons you specified. I wish I could, but I'm not planning to support this (soon), given other priorities and because I personally consider add-path being incrementally deprecated or avoided in ISP practice. I would be happy to learn I'm wrong. Anyway, I'll put add-path on my todo list (to support reading older MRT files), but can't guarantee any timeline.

[netixx]

It's a very nice project :)

In my case, we use add-path to be able to get more visibility on router BGP table in a BGP route-collector scenario.
This enables us to compute backup paths for example.

I submitted a draft PR with a first implementation. I see that it's working in my case (ie. with add-path), but still need to test it more.

[pforemski]

Thanks, I commented in the #3 PR.

BTW, I'll be presenting bgpipe (and bgpfix) at RIPE 88 at the Routing WG (https://ripe88.ripe.net/) in 2 weeks. Feel free to participate (possibly remotely) and ask questions / discuss if you'd like to. Happy hacking :)

[netixx]

I implemented the fixes you suggested.

I won't be able to go to RIPE, but maybe you'll see some of my colleagues from OVHcloud :)
I wish you a good talk, and I will definitely watch the replay!

[pforemski]

Thank you so much @netixx!

I have merged your work and made some changes, primarily enabling best-effort ADD_PATH parsing in case we deal with raw MRT files with UPDATEs only (and haven't seen the session negotiation with capabilities).

I have also modified the JSON format when marshaling ADD_PATH prefixes, to avoid confusion with IPv6 addresses (that also use colons and digits).

Would you have time to have a look at #5 ?