Ive yet got to read the info ISPr.. so no doubt will have something to say about that in a moment.
I've just been reading the ISPr article, and I think it leaves me with more questions than answers...
- Does the increase in latencies that have been seen in the wild really correspond to the fallback diagram? Those would suggest a total of between 16 and 24ms, worst case, but I thought we've had numbers higher than 30ms reported.
- The report suggests that ECI modems can support G.INP downstream, but not upstream. This ought to imply that most cases of extra latency should only be 8ms, coming from fallback of just the upstream.
- The fallback mechanism, combined with the fact that DLM now seems very trigger-happy to apply one of the retransmission profiles to all and sundry, would explain the issues we've seen, providing we can get the numbers to match.
- There is no mention of what equipment is incompatible, apart from that mention of ECI upstream. With the work we've seen from Lantiq and TP-Link, I'm surprised we don't see more indications - but perhaps that will come from a BTW briefing.
- What do the 13ms and 18ms values refer to in the retransmission profiles? My guess is that these values are the new "maximum delay" that determine how many retries are allowed.
- Does the fact that ECI DSLAMs don't support G.INP
upstream mean that, once they're included in the rollout, every subscriber connected to them (including both Huawei and ECI modems) will fallback to a "interleaving low" profile
upstream? Thus forcing an 8ms latency hike upstream on everyone?
- This might answer Kitz's niggle: if the new fallback profiles are actually new ones that act like the "interleaving xxx" ones, but are named internally to something different (perhaps to prevent DLM from cycling back to a G.INP replacement); but separately, the results available to ISPs like Zen don't know what to display for these profiles; we could get the situation we've had reported - no description of the line profile.
- From the mention of the key error statistic - under 1 ES per hour - I think we can assume that this remains an important statistic, even with G.INP active. It tells me that their target, especially for IPTV quality, is still written in terms of the rate of ES.
Time for a beer...
edit: add a couple of clarifications for upstream-only to one point.