Re: [Coin] Charter discussion

Hi Lixia,

Picking on one point.

Please look at draft-farrel-irtf-introduction-to-semantic-routing for our
latest thoughts on what we mean by semantic routing.

Making changes to the semantics of IP addresses is *one* possible approach
to placing semantics in packets to enable routing. But it is not our
objective. I know there are a number of proposals to do this (including WGs
that work on it extensively), but we are not supporting or promoting these
approaches.

Our objective is to look at the consequences for packet routing systems if
they use the different approaches to semantic routing. How does backward
compatibility work? What happens to scalability and stability? What about
security and privacy?

Thanks,
Adrian

PS "Routing" is, of course, a concept used widely at multiple layers.
Sometimes it means slightly different things, but the general concept is the
same. We, however, are considering "packet routing" at layer 3.

-----Original Message-----
From: Coin <coin-bounces@irtf.org> On Behalf Of Lixia Zhang
Sent: 12 November 2021 16:04
To: Marie-Jose Montpetit <marie@mjmontpetit.com>
Cc: coin@irtf.org
Subject: Re: [Coin] Charter discussion

> On Nov 12, 2021, at 6:30 AM, Marie-Jose Montpetit <marie@mjmontpetit.com>
wrote:
> 
> P4 is another language.
> 
> Programming dataplanes (and control planes and distributed systems) is the
goal.
> 
> So now we will have tel-heads -> net-heads -> prog-heads?
> 
> And all in our lifetime!

Agree.  Progress is accelerating :-)

But it seems to me that the word "net" in "net-heads" could benefit from a
bit further classification

1/ communications need networking.

2/ so telecom also built networks, and "tel-heads" in the old days referred
to as "bell-shaped heads", the view that communication is about building a
network that sets up circuits.

3/ really dont mean to put word in your mouth: I assume the "net" in
net-heads means IP networking (as opposed to telecom networks) that we have
today?

If I dare to say this in public (please do not shoot me! close friends have
heard it from me for years:) -- by paraphrasing the old saying, and
clarifying what kind net one is talking about: the "net-heads" in the above
should be called "IP-shaped heads", thinking in terms of shooting packets to
destination addresses. 

4/ the "N" in ICN is also about networking, information-centric networking.

Now back to where this exchange started: yes all kinds of networks need
routing (telecom had routing too, they just did in different ways), so COIN
needs routing support, whether that is in/out the charter is a separable
question. I would just like to remind ourselves that that routing works on
some identifier space, so we need to first clarify what that identifier
space identifiers.

IP address identifies locations (more precisely: network attachment points).
ICN names identifies information--a name can name anything, so the
information can be about anything (sensor data, compute function, Netflix
movie, or a node's MIB or its attachment points...).

As I read from draft-king-irtf-semantic-routing-survey, "semantic routing"
aims to change the semantics of IP addresses.
But I am afraid that the definition of IP addresses cannot be *feasibly*
redefined in fundamental ways --- just look at "IP multicast address" as an
example: that's no longer the address as defined by RFC791, and that
redefinition implied profound changes to the deployed IP base, hence its 30
years of (losing?) battle in deployment.

"The farther back you can look, the farther forward you are likely to
see."--Winston Churchill 

just my 2 cents, Lixia

-- 
Coin mailing list
Coin@irtf.org
https://www.irtf.org/mailman/listinfo/coin