RE: [EFM-OAM] Notes from todays call
I don't think we discussed it to the level of exactly where the
timestamps should go.
I agree with Brian's comments and think that one timestamp per event
OAMPDU is sufficient.
--
Floyd
-----Original Message-----
From: Brian Arnold [mailto:arnoldb@cisco.com]
Sent: Thursday, April 17, 2003 4:54 PM
To: mattsquire@acm.org
Cc: floydg@cisco.com; 'Matt Squire'; stds-802-3-efm-oam@ieee.org
Subject: Re: [EFM-OAM] Notes from todays call
Matt,
I see from my original comment against D1.3 (#978), I suggested
timestamps
in reference to both the event OAMPDU and per event TLV. I wasn't privy
to
the discussions in Dallas, but I think having timestamps in both is
probably a mistake.
If we can make the assumption that an OAM client is populating specific
event TLVs in a given event OAMPDU with counts obtained from the
approximate current time period, and not populating them with counts
from
assorted time periods previous to the current time period (a second ago,
25
seconds ago, etc.), then it's safe enough to send just one timestamp for
the event OAMPDU. Don't need a timestamp for each TLV.
Don't know if others who were present in Dallas found other reasons for
keeping both per-PDU and per-TLV timestamps.
Brian
At 07:05 PM 4/17/2003 -0400, Matt Squire wrote:
>Is there a reason we have timestamps in the individual event TLVs as
well
>as the non-TLV portion of the Event PDU?
>
>Floyd Gerhardt wrote:
>>Re Timestamps:
>>I put together a presentation on why I felt timestamps were important
in
>>Dallas (with Brian Arnold's input). This presentation can be found
at:
>>http://www.ieee802.org/3/efm/public/mar03/oam/gerhardt_oam_1_0303.pdf.
>>Matt unfortunately you were not there that day. The group of us that
>>were there decided to accept timestamps, in order to accurately keep
>>track of events. The presentation shows how the time line can get
>>skewed if we rely only on reception.
>>I was unable to attend the call today, but plan to be on the next
call.
>