Re: IPv6 Formal Anycast Addresses and Functional Anycast Addresses (Fwd: New Version Notification for draft-smith-6man-form-func-anycast-addresses-01.txt)

[Toerless Eckert <[e-mail]>]

On Mon, Nov 04, 2019 at 10:24:24AM +0100, Ole Troan wrote:
> From an architectural point of view, changing addresses midflight, like what ayncast (does/could do) is just a mobility event or changing paths in a multi-homed network.

Agreed.

> We have tried solving these problems at the network layer with among others Mobile IP and SHIM6, and failed.

I thought thre are some good areas of adoption, but yes, there was less
adoption than hoped, but to me that not an architectural failure.

To me the adoption was just about loosing against the L2 competition in
many mobility services.

Or are there specific architectural failulres you see in how IPv6
mobility was done ?

> We need a session layer or a transport layer with it's own connection identifiers.

I don't think we need to reinvent a lot of endpoint functions for every 
transport protocol.

> As far as I'm concerned we have punted this problem to the transport area.

Well, the separation between IP and a good amount of end-to-end
transport functions is questionable, because of the resulting
re-invention of the wheel in transport layer. Alas, the IETF
had been quite incapable to drive more common functionality into
transport layer either with e.g.: failure to establish SPUD.

Cheers
    toerless

> Best regards,
> Ole