Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [EFM] EPON - an observation after the interim

An interesting meeting.

My take on the EPON track is that it walks like an 802.3 passive fiber network (PHY), quacks (MACs) like a token bus or a .16, and has the external appearance, at the 'Real Ethernet ' user ports, of a .1 bridge. The November meetings with .1 and .16 should be very interesting.

May be calling it the full-duplex version of the 802.3 passive fiber standard is the path of least resistance, then take all the arbitration and DBA stuff off to the higher layers (as suggested on more than one occasion at the interim). If the media access control is something other than CSMA/CD then it doesn't fit in 802.3. Existing full duplex 802.3 standards have a null media access layer effectively i.e. no arbitration required, as they are p2p.

A full duplex passive medium for EPON would work with CSMA/CD on the uplink (no need or null on the down link). It would need an extra receiver (low cost) to do the CS and CD parts. Put an overlay access control mechanism (standardised at the ietf) at the bottom of the IP layer, above the MAC, and you can get round collisions. I know it would be more efficient to do it at the PHY, but then it would not be Ethernet anymore. It would be something like .18, or some other existing even numbered group. What was VGAnyLAN anyway, apart from .12?

If the only reason for wanting to call 'things' Ethernet PONs is a marketing issue (heaven forefend as John Major, a UK Prime Minister of little consequence, used to say; or 'you might think that, I couldn't possibly comment', as a fictional PM used to say), then why not claim the Ethernet marketing traction by calling the systems 'Distributed Ethernet Bridges, compliant with IEEE802.1D or 802.1G'. I'll defer to Tony Jeffree and Mick Seaman on semantics of that one. PONs to the specifications as presented at the interim will never quack like Ethernet. They might have Ethernet ports on the outside, but they are not Ethernet MACs or PHYs on the inside. The access control is not CSMA/CD. I may as well try calling my proprietary p2p protocol Ethernet, but it would be better described as the degenerate p2p case of an EPON with a deterministic MAC. How about a full duplex p2p version of .9, with EPON as a multi-point version? (just kidding)

I hope this stimulates some thought, a few smiles, and not too many angry responses. Just my observations as a recent returnee to the 802 process with time to spare on the flight from LAX to Boston.

See you all in Austin for some lively debate.

Best regards

Bob Barrett