Re: [v6ops] Comments on draft-vf-v6ops-ipv6-deployment

[Alexandre Petrescu <[e-mail]>]

I want to thank you for having recorded and re-posted by notes.

Le 17/03/2021 à 11:09, Paolo Volpato a écrit :
> Dear WG,
> 
> Following on Ron’s message, we have listed here the comments shared 
> during the discussion of draft-vf-v6ops-ipv6-deployment  at the past 
> v6ops session, trying to address them.
> 
> Please feel free to comment on; any feedback could be useful to add 
> further considerations in the draft.
> 
> Alexandre**wondered if** 5G can be seen as a business motivator for 
> IPv6, posing doubts on that. Actually, we too think 5G by itself is 
> not the reason for that. From an application perspective, 5G can be 
> seen as a framework to enable new apps or services (or even more 
> users to use them) that may require more addresses. In this sense 5G 
> could help the transition to an IPv6-only network.

In a sense I agree with you.

I think 5G deployments that can be planned in a very near term feature
great opportunities to add requirements for more IPv6.  'More' IPv6
means to have already the known IPv6 in the smartphone and maybe IPv6 in
the GTP, maybe a /63 instead of /64, etc.

Existing 5G deployments - I am not sure they do more IPv6 than 4G
deployments.  Actually I do not know.

(this is speculation:

It might be (from some testing a few monts ago) that some 5G deployment
does more IPv4 in the smartphone than IPv6, possibly with some IPv4
publicly routable addresses that they kept at hand for times to come.

It might be that this existing 5G deployment considers 5G to be
something very new and high performing connections that only very
expensive subscriptions can afford it (e.g. maybe at around 70
euros/month, 'un-throttled' bandwidth if I can say so).  In the past it
was very common that several operators gave IPv4 publicly routable
addresses for the expensive subscriptions, and IPv4 NAT for the less
expensive, 'normal' subscriptions at around 20Eur/month.

On another hand, it might also be that, from an IP addressing
perspective, 5G connections will be precisely like existing 4G and 3G
connections; in that sense they will feature the same IPv6 on the
smartphone that 4G and 3G do.

This might change as we speak, because precisely these days new 5G
deployments are announced where I live (Free and Orange), but I dont
know what they look like.  End of speculation.)

> Michael noted that people active in the enterprise area were not 
> present. True: Nalini contributed greatly to version 01 of the draft 
> while we worked more on other sections in version 02. We agree that 
> more on enterprise can be shared.
> 
> **
> 
> Jordi provided an explanation for the unadvertised IPv4 addresses. A 
> reason why we think more investigation is needed is that the 
> assumptions may all be true but don’t fully explain the trend. In 
> case you missed the pointer, here you find the useful info: 
> https://www.potaroo.net/ispcol/2021-01/addr2020.html 
> <https://www.potaroo.net/ispcol/2021-01/addr2020.html>. The number
> of unadvertised addresses at Dec 2020 is around 50 /8 and it has
> grown in the last 3 years. Even considering the litigations,
> transfers, quarantine, etc. it is hard to achieve that peak and, what
> it is worse, doesn’t explain the constant growth. A large bunch of 
> addresses in quarantine, when released, should cause some 
> fluctuations, but there are no clear signs for that.
> 
> Alexandre also observed the necessity to have a check tool to verify 
> that IPv6 is really deployed when requested. Fully agree. Today it is
> hard to think of that but maybe in a near future that could be 
> possible. Is there any way to involve National regulators in our 
> work?

I think yes, there might be some ways.

 From my part, when I talk to my regulator I understand that they might
not have enough ressources to get involved directly at IETF, but I feel
that they do monitor these discussions and might get help from contributors.

I think that typical regulator activity is very much 'rooted' on
spectrum-, market-equality reasoning around a public good (like when 
Baby Bells were born), and less in IP addressing, which is new, even 
though it is as much a public good as air we breathe is.  Also 'net 
neutrality', and 'services', are attractive domains for regulators, if I 
understand correctly.

The questions I had to regulator about IPv6 and 5G licenses are this:
- it's great that 'IPv6' is a pre-condition to granting a 5G license,
   but why only and too-shortly-said 'compatible to IPv6'?  Why is that
   2-line paragraph so short when there are so many IETF RFCs that could
   be referred to tell what is expected from cellular networks with
   respect to IPv6?
   (the 2-liner in question is excerpt from the "19-1386.pdf" publicly
    available and says:
    "The owner is required to make her mobile network compatible with the
     IPv6 protocol starting December 31st, 2020", my translation of the
     original in French:
    "Le titulaire est tenu de rendre son réseau mobile compatible avec le
     protocole IPv6 à compter du 31 décembre 2020."
   )
   With such a short requirement I can imagine very well an operator who
   might put some IPv6 features in some parts of the network and still
   leave the smartphone on IPv4-only.  Not from a bad intention, but from
   other strong requirements from other parts than from regulator (e.g.
   hardware manufacturers, software writers, and others).

- what is the means to verify that an 'IPv6' pre-condition is respected
   after having granted a 5G license to an operator?  Currently I have to
   work very hard to identify whether this or that 5G deployment tests
   'what is my IPv6 address' a 'yes' on google while the '5G' turned on
   near the 5 bars of signal quality on smartphone screen.  This is an
   end  user tool, but there should be something more authoritative than
   that.  For example, where I live there is a site called cartoradio.fr
   which can offer all the data necessary to tell whether some area has
   5G or not; maybe it should add 'IPv6' to it.  Or maybe some other
   check tool than cartoradio.

   Such a verification tool is important, because it is known that in the
   rush for fast deployment other pre-conditions have not been (or have
   been less) respected, for example when checking how much population is
   covered, or how much area, after a license is granted, etc.

Alex


> Jen commented that having IPv6-only clients doesn’t imply a reason
> to deploy IPv6 (we assume in the underlay). Indeed; let us just echo 
> what Fred said at the beginning of the session about pushing IPv6 
> more in the overlay. Probably we have achieved, for the first time, 
> the end-to-end availability of IPv6 content. We have IPv6 devices at 
> the user’s side and IPv6-based content provided by content providers 
> often based on an IPv6 DC infrastructure. This overlay is where 
> services are offered and where the IPv6 customer base can be 
> expanded. In between we have the networks (the underlay). A good
> move could be to turn them IPv6(-capable) but even if they stay IPv4,
> they can survive (even if it is not the optimal choice). The overlay 
> offers more possibilities.
> 
> Any further comment from the list is very much appreciated.
> 
> Paolo
> 
> *From:* v6ops [mailto:[e-mail] 
> <mailto:[e-mail]>] *On Behalf Of *Ron Bonica *Sent:* 
> Monday, March 15, 2021 2:49 PM *To:* [e-mail] 
> <mailto:[e-mail]> *Subject:* [v6ops] 
> draft-vf-v6ops-ipv6-deployment
> 
> Folks,
> 
> Each week between now and IETF 111, we will review and discuss one 
> draft with an eye towards progressing it.
> 
> This week, please review and comment 
> ondraft-vf-v6ops-ipv6-deployment.
> 
> Fred and Ron
> 
> Juniper Business Use Only
>