Re: [v6ops] Which IPv4aaS mechanisms should CE routers implement first?

[Vasilenko Eduard <[e-mail]>]

Hi Lorenzo,
A few facts:

464XLAT is the must in MBB just because Android and iOS do not support the other 4 mechanisms. Nothing to discuss.

FBB effectively has the choice from DS-Lite->MAP->464XLAT.
DS-Lite because it is supported by the majority of CPEs. MAP because it is the best technically. 464XLAT because it has synergy with Mobile but typically Carrier’s MBB and FBB departments are different with different budgets.

1.       Lw4o6 has a very limited number of deployments.

2.       RFC 8585 is not very old (May 2019). Basement RFC 7084 did not request anything, except DS-Lite. Hence, the majority of FBB CPEs do not support anything else because Retail did follow RFC 7084 only.
Lee Howard has published it here https://mailarchive.ietf.org/arch/msg/v6ops/OAoush6gNQYv1gkPVGmUdUCWgBQ/. It has the reference to Excel on GoogleDoc. DS-lite was leading FBB in 2017 that is predictable.
DS-Lite market share is the biggest but should only decline in %.

3.       MAP-E/T needs mapping IPv4:port_range to IPv6 which results in the requirement of the limited number of big blocks (not XY of /24). It is a luxury for many Carriers. MAP is not much bigger than Lw4o6, but it has the potential because of technical advantages.
The excel would give you the split for 2017. I do not know the exact split between these 3 for now.

Eduard
From: v6ops [mailto:[e-mail]] On Behalf Of Lorenzo Colitti
Sent: Wednesday, November 24, 2021 5:26 AM
To: [e-mail] WG <[e-mail]>
Subject: [v6ops] Which IPv4aaS mechanisms should CE routers implement first?

Hi all,

Is there any data on what IPv4aaS transition mechanisms are deployed by ISPs today? I was approached by a router vendor asking me "how they should provide IPv6-only" and I was looking for data on which mechanisms are most common.

Official IETF guidance in RFC 8585 and 8026 is "do all of them and let ISPs decide, have a nice day", but that's not very helpful to a development team with finite engineering resources and low margins. In particular, it's not helpful to vendors of routers that are updatable and that can thus credibly claim that they can deliver other mechanisms via software update. Those vendors could reasonably argue that it's better for everyone if they launch the most common mechanism(s) first, and then deliver the others in software updates.

It seems like reasonable implementation guidance would be to request OPTION_S46_PRIORITY and parse it, but only initially implement one or two of the mechanisms and leave the others for a future release. If a router did this, what mechanisms should it implement first? MAP-E?

Cheers,
Lorenzo