If I were designing DLM, I'd want an algorithm for ignoring user-initiated events that would work for everybody, not just the expert users who bother to read the small print
Good point.
I should imagine the first problem to overcome with this would be to standardise reason for resync codes. I think we've seen on here that even the BCM based modems can show 2 different sets of codes. iirc it was eric that found the codes used by current BCM 6338/63xxx variants, but even some of those are confusing.
On reflection... using the 15min bin periods is quite a novel approach. It certainly covers the people who shut off their routers overnight type situations. They also appear to have gone one stage further with monitoring for Wide Area Events which should also detect thunderstorms and local power issues.
The only thing its doesnt detect its cant see if the EU is at the other end pulling the power chord or issuing a reboot router command... or even if the modem has flipped its lid and power cycling.
I wonder if the current DLM system could be combined with certain errors such as LoS. The current system tends to look at MTBR and MTBE as two separate parameters and it takes action on either one. I suppose it would be fairly easy to say monitor LOS count within 15min periods which all modems should be capable of. The hard part would perhaps be determining if a LOS happened x seconds before a retrain. Modems only record a count per 15 min bin but not the actual time it occurred to later try tie up with retrain time.
The reason why we have access to more accurate data is because the likes of DSLstats take a reading say every minute and this info is processed on the local machine and we can see that count has increased by 'x' during the one min monitoring period from the current bin. Modems dont do that.. their count unit is in 15 min periods. 15 mins is a standardised measurement across the board for all modems/routers.
