[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Proposed CCAMP recharter



Hi Adrian
 
It's OK you disagree. :-)  I myself would have used the term connection oriented a while back. However I have observed that others have a different interpretation and connection oriented Ethernet is confusing to that crowd. I do believe this a difference between service and network as you point out but traditional Ethernet has a tighter coupling between network and service and the distinction is more subtle. Ethernet packets are not forwarded in a hop by hop paradigm, they follow a VLAN or source based tree always.
 
So rather than put confusing text I prefer what Adrian has proposed. 
 
Regards,
Don


From: Igor Bryskin [mailto:i_bryskin@yahoo.com]
Sent: Thursday, August 16, 2007 9:32 AM
To: Fedyk, Don (BL60:1A00); Adrian Farrel; MEURIC Julien RD-CORE-LAN
Cc: ccamp@ops.ietf.org
Subject: RE: Proposed CCAMP recharter

Hi Don,

I disagree with you on this.

It all depends on definition of connection. If one agrees on the following major properties of a p2p/p2mp connection:

a) single source;
b) zero flexibility in forwarding packets along the connection;
c) deterministic resource reservation

then one can see the difference between, say, Ethernet service (which is an association between source(s) and destination(s)) and Ethernet connection (which is a set of network resources connecting the source to the destination(s)). Ethernet services may be provided by either connection-oriented or connectionless Ethernet networks. Ethernet services, as we know, may be also provided by non-Ethernet (e.g. IP/MPLS) networks).
PBB-TE IMO is all about Ethernet connections rather than Ethernet services.

The bottom line: I think it is correct to have the "connection-oriented Ethernet" clause in the CCAMP charter.

Igor

Don Fedyk <dwfedyk@nortel.com> wrote:
Hi Adrian

Hi Adrian

The description of the term "connection-oriented" does not hold quite
the same meaning in the IEEE as in the IETF. In IEEE, my understanding
is that unicast and multicast services in Ethernet are all considered
"connections" or "associations" when they are part of the same Service
VLAN or Service Instance. So many types of Ethernet services can be
considered "connection-oriented".

Therefore the term "connection-oriented" might be confusing. (The IETF
also considers TCP "connection oriented" even though IP is
connectionless datagram based.)

I would suggest we use something descriptive like "Explicit Route
Controlled Paths", or "Automating Configured Ethernet Paths".

Just a suggestion,
Don

-----Original Message-----
From: owner-ccamp@ops.ietf.org [mailto:owner-ccamp@ops.ietf.org] On
Behalf Of Adrian Farrel
Hi Julien,

> I agree with Dimitri about the use of the term "Ethernet". As we don't
> control IP but MPLS-TE, we're looking into controlling a "connection-
> oriented Ethernet" which isn't really Ethernet and not supported by
> typical "Ethernet switches".

OK. I'll modify the text I just suggested to Dimitri, to say
"connection-oriented"

> On the other hand, (echoing Dimitri again)
> we are also considering Ethernet services over any GMPLS-controlled
> layer.

Agree. But don't think we need milestones.

> I also agree it is useless to rearrange all short-term milestones
(unlike
> the
> charter which has a more long-term value).

I tend to agree, but I suspect that the IESG will find it hard to
approve a
re-charter with dates that have been passed.

Thanks,
Adrian






Pinpoint customers who are looking for what you sell.