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

RE: [802.3ae] Clause 46.3.1.4 Draft 3.1




Jacob,

Bob Grow is the editor for Clause 46 so comments on that clause should be
addressed to him. I'm the editor for Clause 49. 

That said, you appear to be misinterpreting 46.3.1.4. The text looks clear
to me and I'm not sure what is causing the confusion. 46.3.1.4 is not about
data rate control at all. The RS is not responsible for controlling the data
rate. The MAC sends at the data rate and maintains a minimum interpacket gap
between packets. 

The RS has to align the Start character to 4 byte boundaries since the MAC
is unaware of that physical layer constraint and can start a packet at any
time. 46.3.1.4 is about that process. Note that alignment to 4 byte
boundaries does not change data rate. It may cause a packet to be delayed by
up to 3 bytes in order to put the Start character in Lane 0, but that
funtion just shifts 4 byte times of idle that would have fallen after the
packet to fall before the packet. The total amount of idle time is not
changed.

The lines you referenced say: "Note that this may result in inter-frame
spacing observed on the transmit XGMII that is up to three octets shorter
than the minimum transmitted preamble specified in Clause 4, however the
frequency of shortened inter-frame spacing is constrained by the DIC rules."


That clearly says the amount of idle between packets can be changed by the
alignment process. There is absolutely no mention of altering the amount of
preamble.

Neither the RS nor any other part of the Physical layer is allowed to alter
the length of the preamble for 10 Gig Ethernet.

The /T/ is part of the interpacket gap.

Regards,
Pat

-----Original Message-----
From: jacob [mailto:jacob.mathews@xxxxxxxxxxxxx]
Sent: Tuesday, August 07, 2001 4:40 AM
To: Julio.Hernandez@xxxxxx; THALER,PAT (A-Roseville,ex1);
stds-802-3-hssg@xxxxxxxx
Subject: [802.3ae] Clause 46.3.1.4 Draft 3.1


Hello,

Q1.
>From Clause 46.3.1.4 Draft 3.1 Page 270 , line 4 and 5,
it appears that the effective data rate control is achieved by dynamic
preamble shrinkage (i.e either one or two
32bit columns of preamble before start of  ethernet destination
address-as shown below)

 I column             II Column         III Column
   SOF                PRMBL               DA
   PRMBL              PRMBL               DA
   PRMBL              PRMBL               DA
   PRMBL              SFD                 DA
or
 I column            II Column
   SOF               DA
   PRMBL             DA
   PRMBL             DA
   SFD               DA

In draft 3.0 Clause 46.3.1.4 Draft 3.0 Page 251 , line 37and 38,
it appears that the effective data rate control is by dynamic addition
and deletion of IPG columns in between
ethernet frames(i.e the IPG columns will dynamically change to say 2 or
4 cloumns for a nominal 3 IPG
colmun(96bits).

Q2.
Is it mandatory to support effective data rate control only by dynamic
preamble column shrinkage?

Q3
Is it right to consider /T/ (End of Frame Character)as IPG octect?
(Minimum 5 octects IPG infers- one 32 bit column of IDLE and an octect
of /T/ ; refer 46.2.1 line 53)

Please clarify the observations and comment on the same.


regards
jacob