IPv6 support on Virgin media

[dgcarter]

Does anyone know whether (and if so when) Virgin plan to implement IPv6 on its network?

[GreenReaper]

I'm not waiting any longer. Sorry, but this thread has been rolling for almost eight years. Maybe Virgin considers IPv6 to be "just for techies", but it indicates the priorities behind the scenes. Other ISPs may not deliver the headline Mbps you get over cable, but the bigger ones now seem to have their house in order, at least for new customers - and since one made a good offer this Black Friday, we took it. Didn't even bother asking Virgin if they'd match it.

I'll drop by in 18 months to see how things have been going (and if they have anything good yet in 4K). Maybe by then it'll be more than a staff trial.

[Morgaine]

I'm sorry to hear that, GreenReaper.  I'm sure that you're not the only one leaving Virgin Media through their failure to provide a native IPv6 stack.  What's more, customer departures are sure to accelerate the longer that VM delays on this.

After all, IPv6 deployment currently stands at around 41% in Southern Asia, 39% in USA and the Americas, 33% in Western Europe and 21% in Northern Europe [APNIC figures].  Those are colossal numbers, and failure to embrace and support this clear worldwide adoption is business suicide.

Over at the UK IPv6 Council group on LinkedIn, we were linked this very informative blog post about IPv6 Deployment at Dropbox, in which it was hugely depressing to see that the UK didn't even figure among Dropbox's stats for the world's top 10 IPv6-using nations.  Of our "Big Three" UK ISPs, Sky has certainly done their part of the job, so the blame for this failure lies entirely at the feet of Virgin Media and BT management.  As you say, VM customers have been calling for IPv6 deployment for some 8 years.

It's a very bleak situation.  The only faint ray of hope comes from seeing some kind of IPv6-related trialing in progress at VM through APNIC's public IPv6 stats.  Unfortunately, managerial failure to shout from the treetops that IPv6 is coming means that most people won't know that there is any movement on IPv6 whatsoever at Virgin.  Very bleak.

[SlySven]

Just to confirm for the hard of thinking. Those IPv6 deployment figures for Southern Asia, the Americas, and parts of Europe represents sites that operate on IPv6 but that is slightly deceptive as it does not indicate how many of them are dual-stack because it is the proportion that are IPv6 ONLY that we have to worry about and which will be totally inaccessible to VM customers. Now if we can get some sizeable values for that demographic that will be something to be forcefully inserted into the VM group-think that is sticking its fingers in its ears and loudly saying LA-LA-LA to itself...

[Morgaine]

SlySven: The numbers that you're after would certainly be interesting, but we don't have them. Short of asking APNIC or some other big IPv4+IPv6 monitoring site to generate them, we're stuck with what we're given. The current stats do at least have the merit of being statistically very robust, because of the huge population sizes involved.

It's also a bit of a Catch22 to find a list of IPv6-only websites and expect them to influence people whose IP worldview is limited to IPv4, since they won't be able to see them. You'll probably have wasted your time.

The problem is actually worse than that though. The IPv6-only web is hampered by the IPv4 centricism of the last few decades, because the markedly sparse end-to-end user connectivity on IPv4 has created a highly centralized Internet where large corporate sites host the content created by millions of end users. When these corporate sites add IPv6, they don't drop their IPv4 connectivity, so you can't expect the IPv6-only website figures to rise by much to reflect IPv6 usage, at least not in the short term.  The web has been carved up and is effectively owned by the megacorps.

With the unhindered end-to-end user connectivity on IPv6, what the web has become may change if people wish it to, but that doesn't necessarily mean that it will. Companies like Facebook will be trying hard to stop it happening in order to retain their captive audiences. This web centricism is the big elephant in the IPv6 room --- I expect IPv6 denialists to use "Where are the IPv6-only websites?" as a favourite ploy. Don't fall for it, as there are other factors at play.

Although there is little we can do about web centricism, end-to-end IPv6 connectivity will influence all protocols, and many new ones can be expected to appear and in time even to rival the web. It's these new protocols that I look forward to the most, new network abilities giving rise to new applications, many of which won't work on IPv4 because of its broken connectivity. Although the longer addresses of IPv6 get all the press because they're easy for people to understand, IPv6 actually heralds a whole new era in networking, far beyond addressing.

[couling]

This has always been the fallacy of VM's view that "we have enough V4 addresses so we don't need to support v6".  If even a tenth of a percent (1 in 1,000) websites sites that I use become IPV6 only (and thus inaccessible to VM customers) then I'm out of here.

VM needs to remember they have a diverse set of customers.  I am using a lot of Chinese websites.  I'm really concerned that those are going to start to disappear in the next few years, maybe sooner.

[ravenstar68]

Thanks to Teredo people SHOULD actually be able to visit IPv6 only sites anyway.

I say SHOULD as currently it works with any browser OTHER than Google Chrome, due to what appears to be some poor design on Google's part.

I was actually playing with this this morning.  I have turned off my v6 tunnel, and re-enabled Teredo (on most Win 10 setups, it's running by default).

Result with IPv6 only site: testv6.madore.org/

Chrome:      Fail (ERR_NAME_NOT RESOLVED)
Edge:          Connected to page
IE11:           Connected to page
Firefox:       Connected to page

Running a check with Wireshark, shows that Chrome only looked up the site's IPv4 record (it doesn't have one), then gives up.  Whereas the other three check for both IPv6 and IPv4 records.

Note that when I have my normal HE.net tunnel running, Chrome would connect to IPv6 only sites without issue, so this actually appears to be a deliberate programming choice on Google's part, and a poor one at that, being that teredo was designed to favour IPv4 if available but use IPv6 otherwise (correct me if I'm wrong guys).

However I do think Virgin Media needs to pull it's finger out when it comes to IPv6 deployment they've now missed two promised  dates for the commencement of IPv6 deployment.

While one hopes that they don't deploy dual stack lite over here, given the problems it causes with regards to NAT on IPv4, IPv6 rollout does need to happen sooner rather than later.  In fact, the longer they leave it the more likely it is that they'll be pushed into a scenario where they HAVE to use DS-Lite.

Tim

[Morgaine]

ravenstar68 writes:

> "While one hopes that they don't deploy dual stack lite over here, given the problems it causes with regards to NAT on IPv4, IPv6 rollout does need to happen sooner rather than later.  In fact, the longer they leave it the more likely it is that they'll be pushed into a scenario where they HAVE to use DS-Lite."

Those who run their Hub 3.0 in Modem Mode may have noticed that the three IPv6 DS-Lite fields (all empty on mine) that used to be displayed in its Admin->Info screen vanished towards the end of 2017.  It seems fairly reasonable to see this as indicating that they've abandoned their previously assumed plan to use DS-Lite for IPv4 provisioning over an IPv6-only infrastructure.  If so, those who understand the problems that stateless DS-Lite always brings may be able to sleep a little easier for now. 😉

It's no huge surprise, because deploying IPv6 using Dual Stack provides a much less bumpy transition for both customers and the ISP, so if the ISP has the IPv4 addresses to use Dual Stack, that's the way to go.  They'll run out eventually, but in the meantime it'll give them a smoother ride into an IPv6 world.

It's very curious though that the above change in Hub 3.0 web output occurred without affecting its firmware version label, which is still 9.1.116V as before.  I guess the CPE's HTML pages are held in a different part of flash, rather than being integrated into the firmware blob which carries the version tag.  It sort of makes sense, but it's also bad because it means that we don't have a simple way of detecting when Virgin makes a change to the hub's HTML.

Morgaine.

[Anonymous]

Spotted something interesting and unexpected today when running Wireshark against my upstream Virgin Internet interface on my firewall. It seems that VM's Cisco routers are sending out IPv6 Router Advertisement packets. For reasons I don't yet understand these are being ignored by my LEDE/OpenWrt router which is sending out periodic DHCPv6 queries which go unanswered.

The advertisement is for 2a02:8800:f000:2002::/64 and it doesn't have any other information in the RA. From a bit of digging on https://bgp.he.net/AS5089#_prefixes6 it seems that VM aren't advertising this address range (so no connectivity expected) though according to whois it is allocated to them.

Anyone have any clues or seen this before?

The full text from the Wireshark dissection:

Frame 422441: 118 bytes on wire (944 bits), 118 bytes captured (944 bits) on interface 0
Ethernet II, Src: Cisco_6a:80:1a (**:**:**:**:**:**), Dst: IPv6mcast_01 (**:**:**:**:**:**)
    Destination: IPv6mcast_01 (**:**:**:**:**:**)
    Source: Cisco_6a:80:1a (**:**:**:**:**:**)
    Type: IPv6 (0x86dd)
Internet Protocol Version 6, Src: fe80::5aac:78ff:fe6a:801a, Dst: ff02::1
    0110 .... = Version: 6
    .... 1110 0000 .... .... .... .... .... = Traffic Class: 0xe0 (DSCP: CS7, ECN: Not-ECT)
    .... .... .... 0000 0000 0000 0000 0000 = Flow Label: 0x00000
    Payload Length: 64
    Next Header: ICMPv6 (58)
    Hop Limit: 255
    Source: fe80::5aac:78ff:fe6a:801a
    Destination: ff02::1
    [Source SA MAC: Cisco_6a:80:1a (**:**:**:**:**:**)]
    [Source GeoIP: Unknown]
    [Destination GeoIP: Unknown]
Internet Control Message Protocol v6
    Type: Router Advertisement (134)
    Code: 0
    Checksum: 0x7f6d [correct]
    [Checksum Status: Good]
    Cur hop limit: 64
    Flags: 0xc0, Managed address configuration, Other configuration, Prf (Default Router Preference): Medium
    Router lifetime (s): 1800
    Reachable time (ms): 0
    Retrans timer (ms): 0
    ICMPv6 Option (Source link-layer address : **:**:**:**:**:**)
    ICMPv6 Option (MTU : 1500)
    ICMPv6 Option (Prefix information : 2a02:8800:f000:2002::/64)

[Morgaine]

Well spotted, @davefiddes.  Unfortunately it doesn't seem likely that we'll be able to pin that RA to a role within Virgin's IPv6 rollout.  If the prefix were a /56 then we could reasonably guess that it reflects a full IPv6 end-user delegation, but a /64 could have any number of purposes.  Perhaps it's a VoIP endpoint address, and your Hub 3.0 might have grabbed it for VoIP use if VM had enabled your unit for telephony trials.

It's only a guess that such trials are running, but there's a good possibility of it since Hub 3.0 has telephony connectors and APNIC tells us that the 2nd of two periods of Virgin IPv6 activity is still in progress.  To my eyes, what APNIC is seeing looks like a limited customer trial of something involving IPv6.  VoIP would certainly be a nice side benefit of rolling out IPv6 for VM.

VM's IPv6 activity has continued to decline very slowly from its 2nd peak of 8,642 on 2017_12_13.  I'll just show the figures for the last week:

DATE         AS      Users      IPv6-Users  %UKv6
==========   ==      ========== ==========  =====
2017_12_13:  VIRGIN  15,066,456      8,642   0.05 <-- peak of 2nd "trial"
....
2018_01_09:  VIRGIN  15,144,351      7,253   0.04
2018_01_10:  VIRGIN  15,135,864      7,248   0.04
2018_01_11:  VIRGIN  15,128,961      7,060   0.04
2018_01_12:  VIRGIN  15,118,548      7,048   0.04
2018_01_13:  VIRGIN  15,105,046      7,041   0.04
2018_01_14:  VIRGIN  15,091,003      6,952   0.04
2018_01_15:  VIRGIN  15,079,064      6,845   0.04

This decline in IPv6 counts has followed a very smooth curve down from its peak, exactly as happened in the first period of IPv6 activity.  Unfortunately I don't have a hypothesis for why it should be declining, since once IPv6 is rolled out to a customer, there's no clear reason why usage should then ebb away.

If anyone has a better explanation for what's been happening, I'd love to hear it. 🙂

PS. Did Wireshark catch that traffic on the LAN side of a Hub 3.0 running in Modem-Only mode?

Morgaine.

[Anonymous]

I agree that the prefix looks to possibly be an internal service like VoIP. IPv6 is great for doing these sort of private services using a sane address space.

I've not tried doing anything with the RA offered. Suspect this might be a hiding to nothing without a DNS server or some knowledge as to where to talk to.

My capture was of all the Ethernet packets coming through my SuperHub 1.0 in modem-mode.