[c-nsp] full duplex mismatch speed - dynamips

[Adam Armstrong]

On 19/08/2010 17:52, John Neiberger wrote:
> On Thu, Aug 19, 2010 at 10:07 AM, Adam Armstrong<lists at memetic.org>  wrote:
>> On 17/08/2010 23:50, Justin M. Streiner wrote:
>>>
>>> On Tue, 17 Aug 2010, Alessandro Braga wrote:
>>>
>>>> Verify duplex and speed configurations on interface, the rule is:
>>>> autoXauto, forcedXforced. If problem not solve, disable cdp.
>>>
>>> Also, while auto speed/duplex negotiation is fine for user
>>> workstation/PC ports in most cases, I recommend against using it on your
>>> network infrastructure if you can help it.
>>
>> This is horribly terrible advice.
>>
>> Autonegotiation should always be used as default, nailing should be the fix
>> for when things don't work, and where very old devices don't do autoneg
>> properly.
>>
>> Note that for gigabit, autonegotiation is MANDATORY.
>>
>> adam.
>
> Adam, you are my new best friend. I've been saying this for the past
> few years and people still think I'm crazy. I flat out refuse to
> manually configure speed and duplex for someone unless it is
> demonstrated (or I can verify) that a duplex mismatch is actually
> happening or there is some other extenuating circumstance that
> requires it.

It's merely a case of people not changing habits they learnt in the mid 
90s (and new engineers being taught habits by stubborn older engineers)

It's often something I have to argue about quite strongly in a new 
company. I see far more duplex issues in places where they nail duplex 
as habit, because people often forget to do it on one side (or simply 
don't know how to do it on the device in question), and of course, if 
one side is forced and another side is auto, the result is auto side 
drops to half, and you get a mismatch :)

There is of course the odd device which doesn't autoneg properly (like 
PA-FE-TX), but those are almost always old, and are something we should 
be working around, not something we build policy around.

I'm sure there'll be similar stories in 10 years time when people are 
still (somehow) using classful terminology and policies with IPv6.

adam.