Re: Generic anycast addresses...

[Mikael Abrahamsson <[e-mail]>]

On Mon, 3 Jun 2019, Mark Smith wrote:

> https://tools.ietf.org/html/draft-smith-6man-form-func-anycast-addresses-00

I have now read RFC1546 and now I understand some of the confusion. Did 
the concept of "send SYN to anycast address and get SYN-ACK back from 
actual unicast address" ever get implemented anywhere?

Anyhow, the RFC1546 use of anycast never caught on from what I have seen. 
I have been doing IP networking for 25 years and I've done numerous BGP 
anycast deployment for different services, and this just involves treating 
them as unicast addresses (configured on each host in the anycast cluster) 
but that happens to be ECMP:ed to multiple hosts. When one says "anycast" 
to 99.9% of the networking world, this is what they'll think of. This is 
what draft-smith-6man-form-func-anycast-addresses seems to call "informal 
anycast addresses". It's good (and absolutely needed) to make the 
distinction here, to avoid confusion.

Now, if we're going to do the whole "functional anycast addresses", is the 
expectation that it's going to add the example RFC1546 behaviour for TCP 
sessions, thus needing changes to all stateful L4 protocols in the host 
stack? That's a huge change, and I'm pretty sure the security requirements 
section would be huge? If not, what does the whole scoping actually add? 
What does all this information actually add over well-known addresses out 
of regular unicast space that we're already using?

-- 
Mikael Abrahamsson    email: [e-mail]