Re: ipv6nd-02.txt & autoconfig

[Albert Manfredi <[e-mail]>]

[e-mail] wrote:

> LANE did not seem to need any mention of what was inside the NSAPA: nowhere
> in the protocol is any structure assumed and the only requirements that
> LANE imposes are that clients have uniqueness in the complete opaque bitstring.
> Now I appreciate that IPv6 might have some different requirements but it
> would be a noble goal to try to keep any requirements on the internal
> structure of the ATM addresses as close to zero as possible.

Noble, perhaps, from the point of view of IP. Especially if we further
expect to use IP routers between all LLGs (LISs, whatever).

But insisting on useful ATM addresses is noble from the ATM perspective,
to facilitate cut-through, native ATM effectiveness, and so forth.

So overall, it's best to worry both address formats to death.

> BTW, in LANE, a client can use whatever MAC address it wants to to be bound
> to its "primary ATM address" - there is no requirement for ESIs to match
> the declared MAC address. Also, a client can use as many other ATM addresses or
> MAC addresses as it wants (the bindings between them must be relatively
> static of course) although that is probably not relevant here.

I agree. We need something to support autoconfiguration here. LANE is not
relevant.

Bert
[e-mail]