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

Re: Question about Open-loop in XGMII MAC ...

> I have a couple of questions pertaining to Shimon Muller's proposed
> MAC/PHY rate control solution:

> 1.  Are the IDLES added by the MAC to perform rate compensation in
> addition to those added by the RS to perform start-of-packet
> octet-0 alignment?

No, the MAC computes the number of additional IDLE bytes and inserts
them between packets regardless of the behavior on the XGMII and RS.
The RS will perform start-of-packet lane 0 alignment as per Steve
Haddocks presentation from the July meeting. This may extend or shrink
the IPG by up to 3 bytes.

> 2.  Does this imply that the MAC will always in effect be adding IDLE's
> in multiples of four to the data stream, which in turn will be filtered
> out by the PHY?

No, the IPG extension for rate control in the MAC is performed with byte
granularity. The RS takes care of the lane 0 alignment. This should be
transparent to the PHY.

> 3.  Will the 12 IDLE IPG minimum be required once the MAC data
> is mapped into the SONET frame, or is this only a MAC/PHY
> interface requirement?

For packets that are transmitted back-to-back, the Ethernet frames will
be mapped into the SONET frame including the minimum IPG of 12 bytes.

> Fundamentally if the method of rate control is going to be through
> additional null-data being added to the stream, then it is of my
> opinion that the standards committee has an obligation to detail the
> precise mechanism in which this data is to be added in order to
> insure reliable, predictable,  and interoperable performance of the
> link.  To date this detail has not yet been provided.

To date we do not have a standard yet.
The intention is (as always) to provide all the necessary details that
are needed to describe the behavior of the protocol at the compliance
interfaces, without mandating any particular implementation.