Harald Tveit Alvestrand wrote:
But on mobility, I think we blew it.
I don't know if I agree with this -- most users that I know are pretty mobile and certainly wireless. So I think we do have some kind of a "Mobile Internet" already and there's obviously more to come.
I do not want to comment on the specific mobility protocols. But I'd like to suggest that we tend to think of Mobile Internet issues in a too narrow sense. IP layer mobility is just one part of the complete picture. What we really need is an Internet that works transparently, seamlessly, and securely in a world where connectivity is over wireless and changing all the time. And yes, more work will be needed.
Here are some of the necessary components:
o Is our infrastructure ready to provide network access to a large number of users, roaming across different types of providers? Some of you may not be aware that IETF provides much of the infrastructure components that link layers use to grant access, such as AAA. And yes, ubiquitous network access requires that providers get paid for their service.
o What are the protocols that are used at the client end for authentication and authorization? Are we stuck with web-based logins? Or is 802.1X the answer, or cellular network models?
o Does everything work when you combine them together? Does it work with NATs, v4-v6 transition and the simultaneous use of VPNs or multiple interfaces? Do the protocols expect you to transition to IPv6 completely before you can begin deployment of the additional features? Seems like there's a lot of work to be done in this space...
o Do our layer 3 "control" components such as RD, ND, and DHCP provide sufficient capabilities for, say, detecting when an L3 change has occurred?
o Movement performance. There are many factors in this -- I recently counted how many messages one needs to attach to a commercial 802.11 network using IPv6. The result was 24 messages, but I may have missed some. Ongoing work in various IETF and IEEE groups is making this faster, but work remains.
o Transport implications. How do transport layers and applications react to underlying communications path changes? If a mobility or multihoming protocol hides topological movements from upper layers, does it still need to tell TCP to do another slow start?
o Are all of our protocols capable of operating in heterogeneous network conditions?
Many of the above things (though maybe not all) are being addressed by different IETF groups. Just wanted to point out that the issue is larger than Mobile IP.
--Jari
_______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf