[c-nsp] full duplex mismatch speed - dynamips

[Christopher.Marget at usc-bt.com]

Andrew said:

>> On 20/08/2010, at 5:57 PM, <Christopher.Marget at usc-bt.com> wrote:

>> Two mentions of problems with manually configured gigabit operation.
>>
>> Is there really a problem in that scenario?  Shouldn't be.
>>
>> Regardless of what the UI appears to be doing, you can't do gigabit without autonegotiating.  It shouldn't be possible to create a duplex mismatch on a gig link.
>
>The issue is that the interfaces can usually do 10/100/1000.
>If one side is set to auto, and the other to 1000/Full - the side set to auto will usually drop down to 10/Half, as this is usually the lowest common denominator.

Thank you for underscoring my point about additional myths.  Now we have another.

My original point: Duplex mismatches with equipment running at gigabit (1000/full on one end, 1000/half on the other) just don't happen.  Or do they?  I'm open to being corrected on this point, but only with an explanation of how the (required) negotiation has gone wrong.

New myth: /Speed/ mismatches exist.  They don't.  You'll never see a link operating with one end at 100 or 1000 Mb/s and the other end running 10/half.

That's why we call them /duplex/ mismatches, not /speed/ mismatches.

Incidentally, the scenario outlined above will result in a correctly operating 1000/full linkup because both ends (even the one set for 1000/full) are negotiating.

/chris (Adam's other new best friend)