RE: [ethmac] Re: TxMAC-DeferCounter

["NewsGroup" <nbase@nbase.com>]

Hi,

   I agree that there is no reason to add the DeferCounter and the Collison
counter once reach 16 may drop the packet.

one thing you might want to add is partition as if you are connected again
to several station and one of them is "bad" so it keep sending all the time
you would like to clean your buffers and not making a buttle neck all the
way back, but as this is a single host application you might not need it but
for general purpose MAC I would consider adding it.

have a nice day

   Illan

-----Original Message-----
From: owner-ethmac@opencores.org [mailto:owner-ethmac@opencores.org]On
Behalf Of Maik Boden
Sent: Tuesday, October 10, 2000 12:42 PM
To: ethmac@opencores.org
Subject: [ethmac] Re: TxMAC-DeferCounter


Hi,

> If DeferCounter is removed, how this TxEthMAC gets timeout to stop
> attempt process for transmiting a new packet of data ?
This is done indirectly via counting the number of collisions. The 16th
collision causes rejection of the current frame. Because we know about
maximum
backoff times and so on, we can calculate the maxmimum possible time until
the
TxMAC will stop transmission of a frame. It's about 4.000.000 bit-times
what's
equal to 40 ms at 100 Mbps. In respect to the usage of random numbers for
backoff time calculation the "timeout" would ever be less than this value.
But
it's sure that the MAC would never try to send a frame for a longer time
than
40 ms.

If we want to stop transmission attempts within a constant time, we need a
real-timer interrupt because DeferCounter depends on network speed. So I
think
it's better to use collision counting only because the standard does not
require
a timeout at all. But that's what we are looking for!


Regards
Maik Boden