Hello,

I am rebuilding my entire network from scratch.

Apart from Juniper Router on my network edge, all my routers are mikrotik. I'll use OSPF for IPV4 and IPV6.

My questions are:

1. Do you recommend using OSPFv2 for IPV4 and OSPFv3 for IPV6 or should I use OSPFv3 for both IPV4 and IPV6 on same instance? (if RouterOS supports it)

While I only required OSPFv2 for IPV4, I used RouterOS version 5.26, which I find very stable and bug free for what I needed.

2. What RouterOS version do you recommend for my new deployment, considering I'll be using OSPFv2, OSPFv3, Queues, EOIP and maybe MPLS if required.

Thank you very much for any advice you can give.

Rafael

OSPFv3 is specifically for IPv6 and OSPFv2 is only for IPv4.
I don't think there's any such thing as multi-protocol OSPF.

(OSPF isn't the only thing you have to "set up twice" in a dual-stack environment. Firewall rules spring to mind....)

Yes OSPFv3 can be used to carry IPv4 prefixes:
https://tools.ietf.org/html/rfc5838
It's seems mikrotik does not support it yet.

Yes OSPFv3 can be used to carry IPv4 prefixes:
https://tools.ietf.org/html/rfc5838
It's seems mikrotik does not support it yet.

^^ Cisco does

Yes, Cisco does - I haven't played with this myself because my IOS image that I use in GNS3 doesn't support this - BUT, I read their whitepaper on the subject, and honestly, I don't see any difference in running OSPFv3 with an IPv4 address family designation than just running OSPFv2 for IPv4.... because even using OSPFv3 for both protocols, you must still run each one in separate process instances anyway.

It makes sense because if you use a single database to represent both protocols, then you must guarantee that your dual-stack configuration is completely identical throughout the topology because routing decisions are made based on path cost alone. If there is even one link in that path which isn't dual-stacked, then a forwarding decision for IPv4 would be incorrect if it were based on the cost of a v6-only link, and vice-versa.

Okay, so having said all of that, my current position is that I don't understand the internal differences between v2 and v3 well enough to understand why there would be a benefit to running OSPFv3 as a dedicated process for IPv4 routing vs running OSPFv2 as a dedicated process.

What advantages does OSPFv3 offer in routing capabilities that would cause one to choose it over OSPFv2 for their IPv4 routing needs?

Then my next question - are these advantages to adding IPv4 address family support to OSPFv3 in ROS so great that they outweigh the importance of implementing some very basic things you can't even do in ROS at all with IPv6?

My candidates for IPv6 triage in ROS are:

• ability to specify the link-local addresses of interfaces manually
• ability to specify which subnet of an IPv6 address pool gets assigned to which interface when using "from-pool=yes"
• ability to at least learn a default GW link-local address via SLAAC (I've thought of several ways this could be useful even without NAT66 being required)
• ability to specify the IPv6 DNS resolver addresses to announce in RA messages to LAN clients
• NAT66-PT (stateless prefix translation)