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

Re: Link Status encodings


Please see my responses below.

Boaz Shahar wrote:
> Some comments:
> 1) Since packet transmission process and Link Fault event are mutually
> exclusive, it may happen that the RF Pulse ordered set (POS) must be
> transmitted within a minimum size IPG. Such an IPG is composed of either /K/
> or /A/ in the first column, and /R/ in the second. As it stands now, there
> is no place for POS there. As a result, the number of packets or time until
> the next POS is transmitted is not bounded.

I never said that the packet transmission process and Link Fault "event"
was mutually exclusive. They clearly are not. However, Packet
transmission and POS (link fault message) transmission are mutually
exclusive. IPG is not relevant to POS transmission. In link fault mode,
||P|| follows every ||A|| in the 10GBASE-X PCS. 
> 2)A possible solution may be to transmit POS instead of the first /R/. If it
> has a neutral disparity it can be removed from the data stream as a result
> of clock tolerance

The 10GBASE-X PCS ENCODE process generates the code-groups associated
with ||P|| with the proper running disparity on each lane. No insertion
or removal of ||P|| is required by the draft. Since Packet transmission
and ||P|| transmission are mutually exclusive, ||P|| follows every ||A||
and there are statistically many more ||R||s than ||P||s in a Link Fault
stream and any ||R|| may be inserted or removed, there is no reason to
insert or remove ||P|| for clock rate compensation purposes.

> 3)Another way is to choose either /A/ or /K/ for the first IPG column using
> a Round Robin scheme, so /A/ will appear every second packet and so is the
> POS.

See my responses to (1) and (2).
> 4)Yet another way is to say that the MAC/RS layers stop transmission of
> packets (i.e. transmit /I/ columns) as soon as LF is detected. This solves
> the above case, because the DTX XGXS transmits only IDLE pattern, so that
> the distance between two adjacent /A/ (And there fore POS) is bounded.
> However, I think that in this case you cannot ay that "packet transmission
> process and Link Fault event are mutually exclusive".

See my responses to (1) and (2).

> Boaz

Best Regards,
> > -----Original Message-----
> > From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> > Sent: Wednesday, November 29, 2000 9:10 AM
> > To: HSSG
> > Subject: Re: Link Status encodings
> >
> > Boaz,
> >
> > Idle mode is entered whenever any ||I|| (Idle column) is received.
> > Packet (i.e. data) transmission is aborted at this time. The ||I|| may
> > be the result of the beginning of fault message signaling, or the
> > beginning of an Idle sequence following a packet. At the Node
> > A DTE XGXS a link fault condition is recognized, and Fault mode is
> > entered, after 3 RF Pulse ordered-sets are detected. Data/packet mode
> > is mutually exclusive with Idle and Fault modes.
> >
> > Best Regards,
> > Rich
> >
> > --
> >
> > Boaz Shahar wrote:
> > >
> > > A question:
> > > When the DTE XGXS enters IDLE mode: Is it happens only
> > after recognizing a
> > > link fault condition? (3 times RF) or immediately after the
> > first one? And
> > > in this state: If the RS continue to transmit data, the DTE
> > XGXS stays in
> > > IDLE or transmits the data and only changes the idle
> > pattern so that it
> > > contains POS?
> > > Boaz
Richard Taborek Sr.                 Phone: 408-845-6102       
Chief Technology Officer             Cell: 408-832-3957
nSerial Corporation                   Fax: 408-845-6114
2500-5 Augustine Dr.        mailto:rtaborek@xxxxxxxxxxx
Santa Clara, CA 95054