"More intelligence" around default BGP sessions

[jbemmel]

If not specified, the BGP module assumes both ibgp and ebgp sessions should be established:

attributes:
  global:
   ...
    sessions:
      ipv4: [ ibgp, ebgp ]

Would it make sense to change this, and assume no ibgp if the user included an IGP module and no EVPN for example?
Is it possible to infer user intent - or would that become too presumptuous?

I built a simple topology using OSPF and BGP, and figured it might be hard to find how to disable IBGP sessions

[ipspace]

I would assume that in most scenarios if someone enables EBGP on multiple routers and IGP between them, they'd also want to have IBGP between those routers. At least that's how we were building IP networks since the days we stopped redistributing the full BGP table into OSPF in 1990s (https://blog.ipspace.net/2009/05/ios-fossils-ospf-to-bgp-redistribution/)

[jbemmel]

:)

What if one were starting from an enterprise network that is entirely based on OSPF, no BGP whatsoever, and one were to gradually introduce BGP starting with eBGP between different sites. The "full" BGP table is a handful of summarized routes to private address space - not the Internet table.

In other words: Replace OSPF between sites with eBGP

[ipspace]

And that brings us to "do I need IBGP between WAN edge routers?" question. See https://blog.ipspace.net/2023/02/ce-ibgp-multihomed-site/ and https://blog.ipspace.net/2023/02/multihomed-ibgp-details/ for some details