[f-nsp] 1G speed-duplex concept FESX

[Niels Bakker]

* tdonahue at vonsystems.com (Tim Donahue) [Tue 09 Dec 2008, 00:03 CET]:
>Niels Bakker wrote:
>> * techconfig at yahoo.com (Mark Tech) [Mon 08 Dec 2008, 17:04 CET]:
>>> What is the difference/concept behind the master - slave config?
>> http://kc.forticare.com/default.asp?id=3780&Lang=1 came up in a quick 
>> google search, you may find it enlightening - especially the bit about 
>> autonegotiation being required for 1000baseT by the IEEE
> Although there may have been improvements in auto-negotiation, the 
> following statement from the article is untrue:
>
> 'The notion of “auto-negotiation is unreliable” can no longer be 
> substantiated.'

In the context of GigabitEthernet...


> Considering in the past 6 months I have run into 4 endpoints that had 
> auto-negotiation issues, I would argue that statements like this are 
> not definitive and are not necessarily correct.  Given, all of this 
> equipment was connecting at 100 Mb, not 1Gb, but there are times when 
> specifying your port speed is necessary.  I would say that in 99.9% of 
> cases auto-negotiation is reliable, but you may run into times when 
> there are issues.

And then you come up with an example at 100baseTX.

(Have you really hooked up four thousand devices in the past 6 months?)


> (The end points had issues connecting to both Foundry and Cisco 
> switches on different cable paths including a regular patch cable.)

So, broken end points, them being the common denominator.  I've seen the 
same happen; had Foundry investigate the one case I had with that, turns 
out end points did not actually properly implement Ethernet.

The linked article pointed out that for 1000base-X the IEEE mandated 
autonegotiation and interop testing of same.  That's why the claim about 
it working well was made.  Of course prestandard gear won't match up.


	-- Niels.

--