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

Re: how mobile do we want to be

On 22-mrt-05, at 14:05, Dave Crocker wrote:

 It's you and Avri who keep harping about mobility without addressing
 the technical issues that I or others bring up in response.


 There is no evidence that layer 3 mobility
 is even useful or necessary,

that's not a technical claim. 

I never claimed that all my claims are technical.

and the original discussions, a year or so
ago, had clear examples of interest, with mobile phones switcing between
cell phone linkage and 802.11 linkage being one of the more interesting.

You have to drill a bit deeper as mobile phones aren't an application so it's impossible to determine whether they need layer 4 stability on top of layer 3 mobility.

No, but it does need a home address, so you don't get multihoming with
existing mobility. If you want multihoming AND mobility you need strong
crypto to authenticate adding addresses to existing sessions.

1. the 'home address' model is needed in some mobility cases, but not
others. To assert that one is always needed is to presume details about the
problem and solution space that haven't been stated. 

Let me repeat some text from my first message to this list:

Please don't forget: adding a new address in the middle of a session is a security nightmare. The only way this can be done reasonably is with the help of strong crypto (magic PKI dust) or a home agent that is impervious to on-path nastiness such as sniffing and MitM. Obviously, for a good number of applications strong crypto isn't a problem as they already use it today. But mandating strong crypto for *everything* is very problematic for reasons of performance, configuration and robustness. (Let the person who never clicked "accept" on an SSL warning cast the first stone here.)

2. strong crypto is the typical choice for any mechanism that changes
addresses within a transport association.


Not sure I understand what you're talking about.

Are you saying that we need strong crypto and I should get over it?

Multihomed, mobile, simple: pick any two.

clever, but without foundation and certainly not a 'technical issue'.

Let's for a moment assume that all we need for mobility is a mechanism to add new addresses to existing sessions. Can you tell me what kind of architecture would allow this to happen in a way that is secure enough that it isn't the weakest link in a mobile session over IPsec, but also allows for autoconfiguration and doesn't increase computational requirements significantly over the current situation?

We now have the situation that a group of people wants to do X. Another
group of people wants to do X+Y. Does this mean the first group is now
obiligated to do (X+Y), even though (X+Y) is much harder to solve than
(X)+(Y)?

given the trauma caused by a shim-like change to the architecture and given
the very poor history of getting these sorts of changes adopted by the
Internet user community, the answer ought to be yes.

I can go along with that for a bit, because I wish I could have been there when MIPv6 started and have told them that they were going down the wrong path and should have included multihoming support.

The trouble is that the type of input you're providing right now comes at the wrong time. Despite the fact that it's a BOF the shim effort is well underway, and starting from scratch would be a HUGE waste of time, without any reason to believe that a new effort would result in anything better, whatever "better" happens to be. The shim approach has enormous potential for synergy with lots of other stuff, so I'm sure that we're going to revist lots of issues somewhere down the line. There is also a lot of good oldfashioned engineering to do (protocol messages, failure detection, path selection, ingress filtering, the list goes on). Experience so far has shown that these areas can impose unforseen restrictions, so it's very important to get them out of the way before widening the scope.