Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Hari and train-up sequences


Thanks for the comment.  I hope that by now you've received the next
message that I sent :-).  In that message, I reinforced that I am not a
strong advocate of AN, rather that if AN must exist, the link fault
signalling component should be exploited.


At 11:07 PM 11/18/99 -0600, Roy Bynum wrote:
>As a network architect and implementor, I have found that AN used for link 
>and duplex is the first thing that gets turned off when there is a problem
>initiating a link.  Often turning AN off and manually setting the link speed 
>duplex fixes the problem.  This happens even between different units from the
>same vendor.  I don't know why this works out the way that it does, it simply
>does.  AN is a great idea.  Sometimes it works very well, sometimes it does
>not.  Beyond simple link initialization for the fixed full duplex and fixed 
>speed of a 10gb LAN or WAN PHY I do not think that AN should be considered for
>Thank you,
>Roy Bynum
>JR Rivers wrote:
>> If it didn't work, then it may not have been specified correctly.  I
>> presume that most people were able to get 8b10b coding to work.  I would
>> imagine that they also got other components of the autonegotiation process
>> (like link speed, duplex, etc) to work.  If I'm correct, then it seems that
>> the mechanism isn't flawed, but rather its use :-)
>> JR