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

RE: [EFM] Re: OAM - this is Ethernet Bob but not as we know it


I know most delegates to EFM have the basic assumption that any
implementation will include a bridge or a router, and that implementation
specific prioritisation will meet the need. However, this is not a pure
specification from an engineering point of view. If a company chooses to
build a product that has EFM interfaces and does not include a bridge or a
router engine e.g. a non-802 full duplex repeater with an 802.3 EFM
interface and an 802.3 interface, then they should still be able to
implement the OAM standard on the EFM interface. You could call such a box a
transparent non-filtering (non-802) bridge, that IS an implementation issue.
The point being that such a system might meet a market need and that the EFM
OAM standard should not preclude such a system from using EFM OAM.

If the final EFM standard does not define OAM that works with just an EFM
interface (whatever the OAM transport) then I would assert that we will not
have met the terms of the PAR.

I think it would be a good idea to ask Denny for some further explanation as
to how the 'management in frames' frames take priority over payload frames
within the scope of 802.3. I think I remember (seeing or hearing the answer)
that this relied on an implementation specific use of two queues which is
outside of the scope of 802.3.

From what I remember of 802.1D the prioritisation of BPDUs is an
implementation issue. 802.1D is a protocol specification. There is no
definition of transport over which BPDUs run within the 802.1D spec., other
than an assumption that it is an 802 MAC of some type.

My point being that whilst OAM is a protocol within 802.3, and as such the
protocol can be compared with 802.1D, we are also defining an OAM transport.
The delivery mechanism for OAM transport needs to be deterministic
irrespective of MAC traffic load in order to meet the requirements of the
service provider market place. We can argue semantics on deterministic until
our BSE infected cows come home, but I think we all know what is reasonable
and what is not reasonable from the service provider's point of view. They
will test with zero packets and with flood packets, and expect OAM to work
at both extremes.

It could also be said by the purist that 802.3CSMA/CD is non-deterministic,
and therefore the OAM can be non-deterministic. However, if we really want
EFM to be a commercial success then we better be prepared to slay a few
sacred cows, this is Ethernet Bob (Metcalf) but not as we know it. (calf /
cow  stretching it - to 10G and beyond)

Best regards