[Top] [All Lists]

Re: IETF Sub-IP area: request for input (fwd)

2002-12-07 10:59:09

Danny McPherson wrote:
They defined ethernet. It is they who would best determine how to carry ethernet over another protocol and keep current ethernet correctness.

Sure, but what about IP network correctness (e.g., security or congestion control)?
Security isn't an IP issue; it's an IPsec issue. If they use IPsec, then they need to show how current solutions won't work. Ditto for ICMP. etc.

IP doesn't have congestion control. Neither does Ethernet. Ethernet over IP is unlikely to have meaningful congestion control as a result.

Certainly IETF-ers would be useful participants, but keep in mind that transport protocol discussions usually focus at the transport layer (in this case, ethernet) with experts thereof, not at the IP layer.

I'm confused... Transport Layer (as in Layer 4) or "transport layer"[?] (as in Layer 2)? Regardless, being an Internet "user", I'd prefer they be
defined in the IETF.

Running layer 2 over layer 3 is liable to generate this sort of confusion. As far as IP is concerned, ethernet over IP means ethernet is a 'transport' protocol (layer 4).

Not everything spec'd in IP is defined in the IEFT. Certainly they need a protocol number. An assumption is that the IP side cannot reasonably change to support this new use (or at least that it hasn't been shown it won't work), there's nothing left for the IETF to care about, IMO.

That may be the case, i.e., case for a joint group, but this is clearly outside sole-IETF scope per se.

You've obviously observed IEEE member participation in these IETF WGs, hence your comments regarding semantics discussions... And I don't recall anyone saying discard input from individual members or officially from the "original host organization", quite the contrary, actually.

I think the IESG (& IAB) saw value in performing this work in the IETF largely because transporting these protocols (e.g., Layer 2 protocols) over IP networks
was going to happen, regardless, and you could have it only one of two ways:

 o Ethernet (e.g.) PRECISELY specified over IP
or o Ethernet specified over IP in an IP-friendly manner (I'm sure
   the "original host organization" would make attempts, but the
expertise for doing things in an IP-friendly manner clearly resides within the IETF).

Perhaps you can tell us what "IP friendly" means.

Then perhaps there would be a reason for this discussion in the IETF. Absent that, it's an ethernet discussion - ethernet over Frame Relay, ethernet over copper, ethernet over ATM and now ethernet over IP.

Those discussions happened elsewhere, excepting the last. There is clear reason to have that set of discussions in the same place, with the same gathered expertise. IETF members might be interested, and participate, but there's no justification yet shown for IETF hosting this group.