Modified CGAs? [Re: about Wedgelayer 3.5 / Fat IP approaches]

marcelo bagnulo braun wrote:
> 
> El 30/06/2004, a las 17:29, Brian E Carpenter escribió:
> 
>>>>
>>>> CGAs have other downsides for multihoming, too.
>>>>
>>> such as?
>>
>>
>>
>> <chair hat off>
>>
>> I really don't want to start a discussion here about cryptographic
>> strength, but my objection to CGAs in the multi6 context is that they
>> include the /48 prefix in the hash, and that is variable in multi6,
>> which means that the host ID changes when you change prefix. I think
>> that's an unfortunate property because it eliminates some possible
>> tricks in the ID/locator split.
>>
> 
> Well, probably it is my fault for using the wrong terminology, but i was 
> considering cgas as a concept rather than the specific approach followed 
> by the send wg.
> for instance you can consider the cb64 draft or even the ideas discussed 
> between Iljitsch and Jari about including all the prefixes or some other 
> workarounds.
> 
> My point was basically that having a string that it can be used as a 
> valid locator and that it also has a crypto nature would provide several 
> benefits, since it enables a way to prove its ownership and apps can 
> treat it as a regular ip address
> 
> Makes more sense?

Yes, and my first thought on reading the SEND CGA draft was that a very
simple modification to CGAs - not including the /48 in the hash -
would make them suitable for multi6 and still make the attacker's problem
sufficiently hard (2^80).

On another point elsewhere in this thread, my assumption has always
been that the need for uniqueness applies only to {prefix + subnet + CGA}

    Brian

> 
> regards, marcelo
> 
> PS: it is clear that if we are going to use one locator as ULP id, the 
> id will have to change when renumbering, the benefit of this, is that 
> apps still handle ip addresses and that most of their assumptions still 
> hold.
> 
>>    Brian
>>
> 
>