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

[EFM] Changes to 100BASE-X and 1000BASE-X PCS, 10G RS

IEEE P802.3ah D3.1 proposes changes to:
	PCS and PMA sublayer, 100BASE-X.  See 66.1 and 24;
	PCS and autonegotiation, 1000BASE-X.  See 66.2 and 36 and 37;
	RS for 10Gb/s.  See 66.3 and 46.
and associated changes to management.

(There's also a great deal of new, good stuff which is in addition to the current 802.3, rather than having the effect of modifying it.)

The changes may have the unfortunate effect of changing some existing implementations to being non-compliant with 802.3.  They might cause interoperability issues (need more reviewers to establish if they do or don't).


The well established practice of using 100BASE-X silicon with OC-3 like optics, called something like "Fast Ethernet on single mode fiber", which 802.3ah is meant to be standardizing under the name of 100BASE-LX10, now might fall foul of an effectively retrospective change, in which P802.3ah D3.1 says the PCS and PMA should support currently forbidden behavior.

The 100 Mb/s bidirectional link type, standardized as TS-1000 by TTC and as G.985 by ITU-T, uses the current 802.3 PCS/PMA and a PMD compatible with 100BASE-BX10 - a PCS/PMA change would cause fragmentation and confusion.

P802.3ah has objectives:
	1000BASE-LX extended temperature range optics, and
	1000BASE-X up to 10km over SM fiber,
with the intention of bringing standardization with its associated benefits to the widely used "stretched 1000BASE-LX" or 1000BASE-LH; the to-be-standardized PMD type being called 1000BASE-LX10.  But D3.1 says the PCS should support currently forbidden behavior for 1000BASE-LX10, while it says such behavior is to be optional for 1000BASE-LX or 1000BASE-SX.
As 1000BASE-LX and 1000BASE-LX10 are interoperable by design, and
as a particular GBIC could be sold as compliant to 1000BASE-LX or 1000BASE-LX10 while being identical, and
as pluggable optics means that any user can swap between 1000BASE-LX, 1000BASE-LX10, 1000BASE-SX and even 1000BASE-T,
it is impracticable to demand different, unfamiliar, behaviour from the PCS, according to the paperwork associated with a probably pluggable, hot-swappable module.
It is not yet clear whether this causes an interoperability problem (need to understand how link starts up and reacts to faults with Cl.37 autonegotiation on just one end) but it IS clear that this is a market problem, making a retrospective change to 1000BASE-X PCS requirements and tending to fragment the market and/or weaken the validity of 802.3.

The modified 10G RS may cause interoperability problems interworking with current compliant 10G RS (I doubt this last one, but we need representatives from the 10G community to say yes, we have read this, understand the implications, and it's OK.)

At the last 802.3ah meeting we spent some hours investigating these issues, and learnt quite a lot of detailed points.  The committee was not willing to make significant changes to the draft at that time.  The problems remain, and the clock is ticking; we should get to the bottom of the issues, clean up the problems, and progress 802.3ah.

Therefore, I urge particular review of the subclauses mentioned above so that we don't go round in circles at the meeting, 15-18 March.  The comment deadline is midnight EST, Wednesday 3 March.

Thank you,

Piers Dawe