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

RE: Local Fault/Remote Fault

Sorry for getting behind on the Fault email strings, but I had too give
priority to draft review and comment writting for a bit. 


The WIS sends a primitive to the PCS when it is not able to decode data so
that is already covered and it does not need to send all 1s or all 0s. If
there is a signal detect line as from the PMD/PMA to the PCS, there is one
place where a device has fault detection capability without the ability to
send an ordered set. The 10GBASE-LW4 PMA currently has to get SONET/SDH
frame lock before transmitting. It currently sends all 1s if it does not
have that lock. There is one additional potential case. When a retiming
10GBASE-X PMA is attached directly to an XGXS so that the XGXS is really
providing the PCS functionality, it isn't clear that the XGXS receives a
signal detect line becausse it isn't part of the XAUI interface. We will
need to decide whether it is better to add provision for the signal line
there or to require the XGXS to have squelches on its XAUI receive lines.

You are correct about clause 49. The implementation of fault signalling in
clause 48 is currently more complex than necessary. I have submitted
comments suggesting a simpler alternative. If they are accepted, they will
provide for generation of a proper idle with Pulse ordered sets integrated
without restricting data flow.

Pat Thaler

-----Original Message-----
From: Ben Brown [mailto:bbrown@xxxxxxxx]
Subject: Re: Local Fault/Remote Fault

...Stuff from other people deleted

I think the point Mr. Finch is trying to make here is that
some devices, e.g. the PMD, PMA & WIS, don't have the capability
to generate an ordered set. I don't agree with the "all zeros or
all ones" part but, if you replace that with "signal detect", I
would say that part b is very valid.

...Stuff from other people deleted

There is nothing currently in clause 49 which restricts data
flow when it receives ordered sets. Other than decoding, it
knows nothing about the ordered set protocol and leaves that
up to the RS. I understand that clause 48 restricts data flow
after detecting the right combination of ordered sets but I
think that is out of necessity for the generation of the proper
IDLE sequence.