Facebook over IPv6

Ted Mittelstaedt tedm at ipinc.net
Sat Jun 11 20:36:37 CEST 2011


On 6/11/2011 1:56 AM, Gert Doering wrote:
> Hi,
>
> On Fri, Jun 10, 2011 at 05:28:55PM -0700, Ted Mittelstaedt wrote:
>> What technical reason is there that you would run separate IPv4 and
>> IPv6 DNS servers?  The server software can do it so the only reason
>> they would stick that requirement in there is redundancy.
>
> What they want is
>
>   "your unsuspecting IPv4 users only get google-over-v4"  (Resolver 1)
>
> and
>
>   "those users that actually have IPv6 connectivity use Resolver *2*, and
>    that resolver bring you google-over-v6".
>
> If you look from their point of view "if most of your userbase doesn't
> have v6 yet, giving you v6 records will not do any good, but might do
> harm" it makes sense.
>

Of course it makes sense from THEIR point of view.

It doesn't make sense from MY point of view.

If THEY want ME to do something to HELP THEM, then they need to
have requirements that make sense TO ME.

When Google starts paying me money to run my network then they
can have the sort of control over it that allows them to dictate
non-standard "solutions" such as running multiple nameservers.

And with all due respect to the people who have pointed out that
you can run multiple instances of a nameserver on a single physical
box, your entirely missing the point.  Doing it their way means
more complexity on my side, more work I have to do.  I guess I'll
have to spell it out for those folks since they don't get it.

Google is attempting to pressure ISPs to do non-standard things
to make it easier for Google.  They don't get the privilege to
do this unless they are paying those ISPs.

If Google had a public list of IPv6-compliant ISPs that would help
steer customers to me, then I'd consider that more than adequate
compensation - it's not money, but it's free advertising.

But until then, what they are requiring is akin to the extortion
that Disney and ESPN do where thy force ISP's to pay them money
to whitelist their networks.  We don't participate in that, either.

When an ISP on this list submits to Google's requirements your
just encouraging other content providers to make these sorts of
demands that benefit them.  If you let them get away with it,
the others will just decide this is OK too and pretty soon you
will have hundreds of different requirements from different content
providers.  Is that the sort of thing you want on the Internet?

> Auto-Provisioning of the CPEs could be used to give "the correct" resolver
> address to your customers, depending on what service they get.
>

The point isn't how easy it may be to comply.  The point is that
Google has no legal or moral right to do this.

>> And when you virtualize or whatever, you lose the redundancy.  Thus
>> to me your "cheating" and going against the meaning of the Google
>> requirements.
>
> Redundancy is not the goal here.
>
> Personally, I think they should just go forward with IPv6 and stop the
> whitelisting project [or be more liberal about it, and implement an
> automatic mechanism to get resolvers whitelisted] - but then, we do not
> operate a global network, and I have no real idea about the amount of
> IPv6 brokenness in non-european networks.  So there might be indeed
> good reasons to be a bit more reluctant on "global IPv6 turn on".
>

If they have a published research paper that shows that they actually 
did the work to survey this, I would say they have a point.

If they turn off the whitelist for a week and hell breaks loose and
they document this then I would say they have a point.

But they decided this would be a problem before any research.  Thus
they have no leg to stand on.

Fortunately, most content providers on the Internet are not 
sophisticated enough or have deep enough pockets to engage in these
shenanigans.  They will end up doing it the normal way and just
turning on IPv6.  And if Google continues this whitelisting nonsense
then eventually I'll just tell people to use Bing.

Ted

> OTOH, they *could* do it "by region"...
>
> Gert Doering
>          -- NetMaster




More information about the ipv6-ops mailing list