Re: [arch-d] ETSI launches new group on Non-IP Networking addressing 5G new services

[Dino Farinacci <[e-mail]>]

Right, so you want new forwarding features. But let’s pop up to a higher level discussion. I think what is more important than technology at a single layer is that you have to get it deployed with useful features at other layers. A holistic approach. And if you want to run it over the Internet, your new features may or will not work well.

IPv6 shows up how a simple protocol takes a long time to get deployed. And that the applications that run over IPv6 could also be run over IPv4.

I’d argue that when IPv6 was developed and if we created QUIC at the same time, people would argue there are more useful features that merit the change. IMO, I think its still not enough.

Dino

> On Apr 9, 2020, at 3:23 PM, Toerless Eckert <[e-mail]> wrote:
> 
> On Thu, Apr 09, 2020 at 03:06:00PM -0700, Dino Farinacci wrote:
>>> Right now we only have the disucss whether somehing works in forwading plane
>>> individually in each protocol new field discuss, and that does not give us
>>> good coverage. It also raises the risk that someone who does not like a
>>> protocol uses the "not possible in forwarding plane" as a scapegoat
>>> not having to say no on a protocol due to otherreasons (company policies etc..).
>> 
>> Let me try to be specific on your point. I could be completely misunderstand you. Here are some basic statements:
> 
> You make a lot of good observation to discuss below, but my comments
> where not meant to be specific to ICN. Think about something
> else, like BIER, or any other interesting new forwarding plane
> features. Think about a vendor who might have stood up and 
> said "BIER is nogo as my HW does not do it", but suspicion
> could be that he just does not want to see devaluation of
> investment into existing multicast forwarding plane.
> 
> If instead we had a broader agreed upon understanding of what
> forwarding plane functionality we could expect upfront, this
> could not happen. Think about defining some reference thats
> maybe P4++ (we know its too limited, we don't know what an
> acceptable ++ is). So you demonstrate that you can do your
> forwarding plane feature in P4++ and that should settle the
> "acceptable for forwarding plane in IETF" argument.
> 
> Just as a very rough idea on potential goals for forwarding
> plane architecture.
> 
> Cheers
>    Toerless
> 
> P.S.: If you want a discuss about ICN/NDN i am all game. Still trying to
> figure out its benefits over PGM with DLR (kidding).
> 
>> (1) IP is a robust network layer that has proved to work for many applications.
>> (2) ICN/NDN is a new data-plane paradigm that is designed for non-conversational flows.
>> (3) It is not clear yet, if an ICN forwarding plane is faster than a IP forwarding plane. And it may not matter.
>> (4) ICN can do source authentication for each packet. That is a nice feature at some cost.
>> (5) IP cannot do packet source authentication.
>> 
>> So you might follow to say that IP can???t do source authenticaiton and therefore IP doesn???t have enough coverage for new features.
>> 
>> Is that what you are claiming?
>> 
>> Dino
> 
> -- 
> ---
> [e-mail]