MarcoH wrote:
Dear colleagues,
Please see a new article on RIPE Labs contributed by Stephane Bortzmeyer:
How Many RIPE Atlas Probes Believe They Have IPv6 (But Are Wrong)?
https://labs.ripe.net/Members/stephane_bortzmeyer/how-many-atlas-probes-beli...
(private contribution to this list as "somebody who cares")
Thanks for that!!
This seems to highlight the problem that "Happy Eyeballs" was invented to hide: clients who think they have connectivity but in reality don't.
Now I appreciate HE as the customer friendly approach in suppressing the negative effects (timeouts) of this, but if the problem is as big as you say it is, namely over 10% this is a very nice and very big "disaster waiting to happen."
What happens when the first true IPv6-only services appear on net, that is 10% of your install base calling the support desk (if you're lucky) or taking their revenue elsewhere?
Which leads to some questions:
-Is 10% really a representative figure or do RIPEatlas probes have a higher chance of living in some experimental (and potentially broken) network. -Should there be some effort undertaken in finding the cause of this brokenness and getting it fixed.
Quick thought: make the Atlas software "clever enough" (unless it already is :-) ) to decide on its own about having IPv6 to the Internet or not. Then, on the general overview map (RIPE Atlas - Active Probe Locations) add some indication whether it is working or not. That could be as simple as adding a green tickmark or red X next to lines of IPv4 Prefix and IPv6 Prefix (in the probe info pop-up). Or change the colour of green to amber is v4 or v6 "should" work, but doesn't, as warning indication between green and red? That would be for public consumption. A similar approach could be taken for the individual probe herder's "My Probes" summary page, improving the granularity of "Status" Connected in green to be specific about v4 and/or v6. That would be for privae consumption.
And ultimately, is there anything the IPv6 WG should or could do?
Marco
Cheers, Wilfried