RE: (Another) Question Regarding Open-Loop Control Mechanism
Boaz,
I disagree with you conclusions. We have allowed the RS to produce a
minimum gap of 9 octets, but we do not allow it to produce such gaps
continuously. The minimum gap output by the MAC is 12 octets and the
minimum average gap put out by the RS is also 12 octets. By inspection
of the table at the end of haddock_1_0700, one can see that the worst
case is that 1 in 4 packets is transmitted with a gap greater than
12 (15 octets to be specific) and the remaining 3 packets are transmitted
with a gap of 11 octets. This situation will occur when the MAC
continuously sends packets of length mod 4 = 1 with minimum IPGs.
Since the worst case clock skew requires dropping a set of 4 octets
for every 10.67 max size packets transmitted, then it would be
possible to specify that the minimum IPG is 9 both as transmitted
from the RS and as received by the RS. The two can be equal as long
as the RS is limited in how often it can send gaps of less than
9 bytes. Such a requirement would increase the amount of buffering
an elasticty buffer needs over the threshold to 12 * (n +1) bits where
n is the number of non-synchronous sublayers between the transmitting
RS and the sublayer because worst case only every third packet allows
an opportunity to delete an /R/.
I don't believe that lengthening the IPG out of the MAC is an option
that would gain support. Therefore, I believe we have two options (the
first being the same one Boaz suggested though I've reworded it):
1. MIN IPG at the RS transmitter is 9 octets (As is today), and at the RS
receiver is 5, and allow any unit at the middle to remove /R/ column.
For non-8B/10B sublayers allow deleting idle when at least 5 idle octets
have been transmitted.
2. MIN IPG at the RS transmitter and at the RS receiver is 9 octets, and
only allow a sublayer to delete idle when the resultant IPG will be at
least 9 octets.
I'm inclined toward the first option because the 2nd one reduces by
a factor up to 4 the opportunities for clock compensation, but it does
mean we are letting the received IPG get very short.
Pat
-----Original Message-----
From: Boaz Shahar [mailto:boazs@xxxxxxxxxxxx]
Sent: Thursday, September 07, 2000 1:07 AM
To: 'pat_thaler@xxxxxxxxxxx'; Boaz Shahar; bebrown@xxxxxxxxxxxxxxx;
stds-802-3-hssg@xxxxxxxx
Subject: RE: (Another) Question Regarding Open-Loop Control Mechanism
Pat,
> get as short as 9 bytes. I believe it would be possible to write the
> standard so the gap is never allowed to shrink below that or
> to allow it to
> shrink to something smaller.
>
The standard has to refer to two parameters regarding the size of IPG: 1)
The minimum size of IPG during transmission (i.e. while packet is output
from the MAC). 2.The minimum size of IPG during reception (i.e. when it goes
to the receptor MAC).
I think that the last discussion leads to the conclusions that:
1.They cannot be equal. If they do, you cannot do clock tolerance
compensation
2.As bigger the difference between Min IPG size (Tx) to MIN IPG size (Rx) -
System reliability in terms of clock tolerance immunity is better
We may choose one of two options:
1.Either say that MIN IPG during Tx is 9-12 (As is today), and during Rx is
5-8, and allow any unit at the middle to remove /R/ column
2.Or go back to the original proposals with MIN Tx IPG 13-16 (/A/, /K/, /R/)
and 9-12 during reception, an allow any unit to remove an /R/ column
Boaz