[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Mutli6 meeting in Vienna

On woensdag, mei 14, 2003, at 09:34 Europe/Amsterdam, Kurt Erik Lindqvist wrote:

"The Architecture of End to End Multihoming"
<draft-ohta-e2e-multihoming-*.txt> has been giving an overview
on so-called host based solutions in a way not specific to LIN6
before the WG was created.



Thus, I'd like to have a chance of presentation.





As I said yesterday, we will not have time at this meeting for all the 
proposals to present.



Note that the above isn't so much a proposal as discussion with some 
conclusions about what a solution should look like.




Instead we will try and have an overview of solution classes. Most 
work seems to be done with loc/id separation and Christians host-based 
solutions. For the time being this is where I think we should spend 
time at the meeting.



I've been trying to come up with a useful approach for this for a 
while. We really have to get this right because if this turns out to be 
a huge shouting match or nap time, it's the end of multi6.



How about this: if everyone says what proposals they feel are worth 
working on. Then we can discuss those proposals. It seems likely the 
proposals that many people want to work on have some promise. I think 
this is a good way to get around "this proposal sucks" "no it's great" 
type of discussions. The other advantage is that it limits the number 
of "active" proposals significantly. For instance, my geo stuff doesn't 
require work but simply adoption  :-)  so there is no need to go over 
it. Mobility and HIP have their home elsewhere, so those don't have to 
be discussed either, apart from maybe send an annotated copy of the 
requirements to those groups to make this stuff work well with 
multihoming.



So that leaves the whole locator/identifier thing. A few questions 
about that:



- on what layer do we want to do this?

- can we break applications?

- can we break transport protocols?

- can we break autoconfiguration?

- can we break IP?

- can we break IPsec?

- if we can't break something, what are the alternatives?

- where can we add state?

- can it be done without state?

- do we need a discovery protocol?

- do we need a negotiation protocol?




Unfortunately we have some catching up to do and having all proposals 
presented would take up very much time, without us moving on.


Everyone should read the drafts.