Building network automation solutions

9 module online course

Start now!

How widespread is IPv6 on mobile phones?

We had great fun listening to Christian Gotare from Ericsson during the 3rd Slovenian IPv6 Summit (program description as translated by Google Translate). He made numerous very strong statements about the (in)abilities of application programmers (watch his presentation ... it starts at approximately 0:42:00) and concluded his presentation with a live demo: he accessed Facebook through IPv6 from a Nokia phone running Symbian.

We all found the idea of an Ericsson guy doing demo with a Nokia phone hilarious and I thought he wanted to demonstrate that even Nokia could get it done ... until one of the Slovenian mobile operators described the problems they’re facing when trying to deploy IPv6 in their mobile network. According to his short update, only Nokia has reasonably priced mobile phones supporting IPv6.

You will find an overview of challenges faced by Service Providers rolling out IPv6 in the Market Trends in Service Provider Networks webinar. Core network design and implementation guidelines are described in the Building IPv6 Service Provider Core webinar.

I always thought that most mobile phones today support IPv6 (after all, the explosion of IP connectivity on smartphones is one of the primary reasons for IPv4 address exhaustion), but that seems not to be the case. Android and iPhone OS 4.0 supposedly run IPv6 ... but only on WiFi.

Can anyone share more information about other vendors in the comments? Thank you!


  1. indeed this is the case, and this is the major show stopper. More than that, most mobile phone makers don't even have IPv6 support in their roadmaps.... (most saying they are busy with other stuff to implement)
    Only a couple of Nokias now support IPv6, and this is pretty much it.
    The situation is not better for wireless CPE devices (3G routers, dongles etc.)
  2. I presented this stuff on RIPE60 meeting, also show stoppers:

    I asked Cristian, why he presented it on Nokia phone, despite he is from Ericsson; the answer was "well, Sony Ericsson is totally different company, we have no internal access to it and therefore for us it's just another phone making company"

    Brilliant. Cristian is coming to Slo in next couple of weeks, I need to catch up with him and have further discussion...
  3. I don't understand why IPv4 exhaustion has been caused by the amount of mobile phones. From the Cell access pov, they are in my experience, using private addressing. From a wifi connection then it is just the same a a laptop and will get the wifi provider ip (also seems to be normally private addresses.)

    Why not whois mobile operator AS and checkout the supernets advertised compared to fixed line.. for a more scientific answer? 8-)
  4. It's true that many mobile operators use private IP addresses in their access networks and large-scale NAT to connect to the Internet. I cannot judge whether that's a deliberate design decision or a side effect of external forces (for example: nobody would give them as many IP addresses as they claimed they require), but I would suspect that no SP would happily and willingly incur the burdens of large-scale NAT.

    Note: it's possible that the mobile operators would be less hostile toward NAT, as they usually connect IP hosts to the Internet, whereas many CPE devices in landline networks already perform NAT (and double NAT is worse than NAT ... whether logaritmically or exponentially is a matter of personal opinion).

    However, unless we want to partition Internet into many walled gardens connected to the public highway by Large-scale NAT, we need to deploy IPv6-like technology - not because it's fancy, but because the number of required publicly-accessible access points (if you subscribe to the end-to-end Internet idea) exceeds the remaining available address space. Most of that growth comes from residential customers, be it landline or mobile. If you can believe industry analysts (some of them came under a bit of a scrutiny lately), the majority of the growth will be in the mobile space. And yet, neither low-end CPE makers nor mobile phone industry (Nokia being a notable exception) see any need to invest into IPv6.

    Last but not least, I love your idea of getting a more scientific answer. The "only" problem I see is that many large operators provide both landline and wireless services, so it's quite hard to figure (from the outside) what they're doing internally. Any ideas?
  5. So far IPv6 is just talked about.., more and more as the years go by.. One hopes when they do it they get guys like me to design it (experience of ipv6 transit design) rather than our super cheap no knowledge cousins ;-)....

    Regarding the initial design idea. The initial mobile design philosophy was indeed walled gardens, this was because the mobiles wanted to be the content providers.. I had many arguments, as I, coming from fixed line internet, had noted the failure of this model with Compuserv & AOL...(remember?) Anyway I guess that's as good a reason as any as to why NAT has become embedded in mobile operator culture.. Additionally, they also have big content zones where they reformat web pages/compression and other blackbox mobile techniques... imo they are just now beginning to see themselves as part of the Internet at large...

    As an aside, In the end, I got told I am not commercial enough to understand, (I used to get told similar with MPLS applications in fixed lines also ;-) )

    Finally, regarding AS, i think you will find enough mobile operators have their own AS number/PI address space to do a sample big enough. Moreover, a lot of mobile operators are run completely independently per country and so have their own AS per country in many cases. (Just to make it a bit more difficult :-) )

    I take your point re: combined fixed/mobile but in my experience the merging AS projects haven't started yet -- RFC 4364 Section 10 -- . (Lower priority perhaps?) -- Would love to know if my last statement is true across everyone's experience.

    Part of 3gpp standard means distinct separation of 3gpp interfaces this appears in my exp to be done with MPLS VPNs, so a slightly different method of using them then your normal provider customer MPLS VPN Service.. So something of a philosophical architecture challenge but going off topic now into fixed mobile merger howto -- perhaps a future Ivan blog topic?
  6. Hi,

    All S60 phones from Nokia supports IPv6, all N8xx/N9xx supports IPv6, Ericsson Mobile Broadband Modules supports IPv6 (Laptops from IBM, HP, LG, etc.......), Qualcomm also supports IPv6 fully on the platforms they manufacture. And Yes, it's kind of funny to use a Nokia as an example in the demos 8-)

    Again coming down to the baseline, do there exist a fully fledged NAT64/DNS64/ALG64? Otherwise You'll have issues with the communication with the world.

    We are today running streaming, etc... .over the mobile core and have this far only had minor glitches with special applications. The issue with keep alive is still an issue, due to "bad" programming (Needed to be able to traverse over NAT44 devices, not the programmers fault).

    Best regards
  7. Hi,

    The numbers of devices on mobile operators is an issue. As an example we can use a Chines operator. One operator have approx. 400.000.000 users and China in total have approx. 9.000.000 IPv4 addresses. This means that even one operator have an over subscription of the total address space with 40:1. Not a very good situation. If we look in to the american market, they could probably live with current address space. But the problem is that they can't consolidate the existing addresses to one bank and redistribute the addresses. Both political internally (PTT's as an example), and also a fight between fixed and Mobile on operators.

    To get end reachability we need to go for a new address space, in this case IPv6. And live with the fact that NAT64/DNS64/ALG64 are needed for a time of transition.

    Best regards
    Christian Gotare
  8. Thanks for the exhaustive list.

    There's no fast NAT64 that I'm aware of (DNS64 is a smaller problem, patches for bind should work just fine). The open-source code is a joke (oops, proof-of-concept); as Jan will gladly tell you, it crashes after being loaded by more than a few clients.

    Cisco IOS can do NAT-PT in software switching path (and let's assume NAT64 is the sensible subset of NAT-PT), which should bring you to 20-30 Mbps on high-speed platforms. NAT64 on CGSE (service blade for CRS-1) is in the pipeline (scheduled for this calendar year), the implementation on ASR 1000 will probably appear after that.

    Juniper supports NAT64 (actually, it's yet again NAT-PT) on their firewalls. I have no performance figures.

    Best regards
Add comment