Re: [Int-area] Using ISO8473 as a network layer to carry flexible addresses

[Alexandre Petrescu <[e-mail]>]

Le 03/03/2021 à 08:46, Carsten Bormann a écrit :
> On 2021-03-03, at 03:42, Liguangpeng (Roc, Network Technology 
> Laboratory) <[e-mail]> wrote:
>> 
>> backwards compatible
> 
> I would suggest not to center the discussion on terms that are so 
> overloaded that they are essentially meaningless.
> 
> Some believe “backwards compatible” means “does not break existing 
> applications”. That is a low bar (but not trivial either).  It does 
> not imply any form of deployability in the sense of actually making 
> the scheme work.

I agree.

The 'backwards compatibility' concept has many facets.

I could say that using an ISO network layer is backwards compatible in
that it does not break existing IP network layers.  The EtherType in the
layer below makes sure there is a right distinction.

But, has this been tried in practice against many systems?  Have people
sent ISO network layer packets to computers that only implement IP and
that might not have an EtherType value for ISO in their link layer of
their stacks?  What happens them?  Maybe it is just an error reported,
or maybe the system segfaults outright.

This can be looked up closer.

> Some believe “backwards compatible” means “seamlessly integrates 
> existing infrastructure” (maybe with the exception that certain 
> benefits do not accrue until that is upgraded).  That has
> essentially been the basis for all major innovation that has happened
> in the Internet, with the exception of IPv6 (and you know how long
> that took and still is taking in some parts of the world).
> 
> Terms like “backwards compatible” are, however, useful for derailing 
> the discussion if that is one’s intention, because quickly the 
> discussion will be about the terms and no longer about the subject.

If one is to derail discussion, one would also point to 'future proof'
aspects of newly designed protocols.  Designing something today that
might not work in the future is not advantageous.

Putting an IPv6 implementation in a computer that only does IPv4 makes
sure that computer is future proof for when IPv6 arrives.

But can one say that putting an ISO network layer in a computer that
only does IPv4 and IPv6 is future proof for when ISO arrives?  Is there
a plan for deployment of ISO network layer?  Like IPv6 had a plan with
the 6bone, 3ffe prefixed addresses, and the 'island' concept?

Alex

> 
> Grüße, Carsten
> 
> _______________________________________________ Int-area mailing list
> [e-mail] https://www.ietf.org/mailman/listinfo/int-area
>