Well one problem with that is that a port configured to auto-negotiate can only link up in half duplex the link partner is configured as fixed speed. This is called parallel detection. The phy is transmitting fast link pulses (FLP's). On the receive side it can detect FLP's, MLT3 coding for 100mbps, or Manchester for 10mpbs. If the phy receives MLT3 or Manchester it links up the appropriate speed and half duplex, this is required by 802.3. If a port can only do auto-negotiation it can never come up in full duplex when it parallel detects. So if a link partner does not support auto-negotiation but it does run full duplex you could never fix the duplex mismatch. I don't know if this was the original intent behind the way the speed control works, but it would be an issue.
Regards, Jim -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of John Neiberger Sent: Friday, August 20, 2010 9:38 AM To: Mikael Abrahamsson Cc: [email protected] Subject: Re: [c-nsp] The myths of autonegotiate vs forced On Fri, Aug 20, 2010 at 2:18 AM, Mikael Abrahamsson <[email protected]> wrote: > On Fri, 20 Aug 2010, Peter Rathlev wrote: > >> On Fri, 2010-08-20 at 09:34 +0200, Mikael Abrahamsson wrote: >>> >>> In case of having to force, do use "speed auto / duplex full" if your >>> equipment supports it. >> >> And on a side note, if one needs to provide less-than-linerate speeds >> (e.g. a 100 Mbps customer on a gigabit interface) it's much better to >> use "speed auto 100" than "speed 100". Not all devices/modules support >> it (by far), but it will give you the benefits of autonegotiate while >> still "locking" the interface at a lower speed. >> >> The 3560G supports it, most 6500 modules seem not to AFAICT. Let's see >> if 100 Mbps disappears before this will be generally implemented. > > I feature requested this (why stop doing autoneg just because you select a > single speed/duplex, there is no downside to still announcing your > capabilities), the response back was the usual "you're the only one asking > for this"-answer. In my opinion, this is how things should behave by default. Fast Ethernet devices often check to see if they have an autonegotiating link partner and will fall back to half duplex if they don't detect one. For the life of me I can't understand Cisco's reasoning for disabling it completely when the settings are manually configured. It costs you nothing to leave Nway enabled, and it regularly causes problems when it is disabled. What could their logic possibly be? There are two options and they picked the worst one. I can't think of any rationale that holds up to reasoning. _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
