[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: now OT re: cisco-autoneg-sucks; WAS Re: xl(4) opinions



On Friday 07 March 2003 07:31 pm, Brad wrote:
>
> If doing autoneg then both ends should be auto or both ends should be
> hardcoded. Mixing auto and hardcoded is wrong and will guarentee you
> have duplex mismatches.
>
	This assumption falls apart if one end is not capable of being managed.

Case in point:  Cisco 1548 (Off the top of my head, don't remember exactly) 8 
port 10/100 UNMANAGED switch.

Duplex and speed was required to be forced on xl(4) and fxp(4) series cards 
(hell, any card) to get a decent throughput.  

Other end was not configurable, as again, it's not managable.

And it isn't wrong.   And does not "guarentee you have duplex mismatches."

Again, off the top of my head, quite a few of the newer devices just drop 
items from their capabilities table when they are forced.

The problem with some (all?  I don't know) Ciscos, IIRC, is that they 
continously try to negotiate speed and duplex.  If the other end (as part of 
the Auto-negotiation protocol) only reports one, it'll stay with that one.

Again, this may be a YMMV thing, but forcing speed/duplex on just one end is 
sometimes the only thing you can do.

It also may be limited to Ciscos.  I certainly have no problem with my Hawking 
and Linksys units in the field not being able to autonegotiate.

I think we can put this particular dog to bed.

-- 
Joseph C. Bender
jcbender (at) benderhome dot net



Visit your host, monkey.org