Re: Persistent or opportunistic IDs

[Iljitsch van Beijnum <iljitsch@muada.com>]

On 7-mei-04, at 18:04, marcelo bagnulo wrote:

> Yes, and additionally, i guess we have to understand why, if the stable
> identifiers are the long term goal, are we adopting the transition 
> solution
> in the first place.
> I mean, any solution will imply the modification of all the hosts, so 
> why is
> that an intermediate solution is adopted instead of the definitive one?

It makes little sense to implement a solution and then replace it by 
another that's equal or better in every way. However, it may not be. 
For instance, the fist solution may provide better backward 
compatibility than the second one. Also, the second one may be 
"heavier" because it has better security. In situations where a high 
level of security isn't imporant, it could be preferable to use the 
simpler solution.

With two solutions there is less need to compromise. However, I'm not 
advocating building two completely independent solutions, what I'm 
thinking is more along the lines of having two modes of operation: the 
mandatory to implement mode is simple and light weight, but may not 
solve all aspects of the multihoming problem. The more advanced mode 
would be optional, and solve all the remaining problems, but probably 
at some complexity, overhead and compatibility expense.