Slow WiFi with Android Marshmallow & IPv6?

[Ted Mittelstaedt]

On 4/27/2016 12:37 AM, Bjørn Mork wrote:
> Ted Mittelstaedt<tedm at ipinc.net>  writes:
>> On 4/26/2016 1:37 AM, Bjørn Mork wrote:
>>> But let's face it: name-server config is not something that interests a
>>> large group of end users. Any feature which is not part of the default
>>> OS installation is not worth considering at all.
>>
>> is that a fact?  I suppose you don't use SSH then to access your
>> command line servers and devices.<eyeroll>
>
> I meant to imply a name-server config context.  I.e.
>

Understood.  I didn't mean to imply that ISP's should tell
end users to run RDNSS on Windows.  But I did want to dispel any
talk that it was impossible.

> It doesn't matter much what
> joe-anal-it-department-guy does to make Windows work on his DHCPv6 free
> network either.

That I disagree with.  Windows is built for those IT departments.  The
end users on the Internet are basically trying to make a PC operating 
system that is sitting there wondering "where the eff is my domain 
controller" work.  What joe-the-anal does and what he forces Microsoft
to do really does make a huge difference to those ISPs we were talking
about and to those end users.

In the Microsoft world the home end users are the cling-ons on the
hind end of the business market.  They get whatever dribbles down
off the business table.  Windows isn't built for them.  It's built for
joe-the-anal.

>  It's not like it would
> help much if new Windows or Android versions suddenly supported the
> other protocol.  End users don't upgrade.

When Microsoft is handing out Windows 10 for free - yes they DO upgrade.

If I had a dollar for every time I told a customer "sorry sir but you 
can't access Ebay anymore until you run a newer web browser that 
supports the stuff they are doing on that site" only to have him tell
me "Well I'll head over to Best Buy and see what they have on sale I
guess I am going to have to replace my Windows XP system"  I'd be a 
wealthy man.

>  In the Android case, they
> often don't even have the option.
>

That is handled by "no you can't run the latest Angry Birds with that
old of a phone" then they run out and get a new one.

Yes we have this Google vs Microsoft issue right now - in 2016.  But
there _IS_ an upgrade cycle and both MS and Google have experts who have
figured out how to drag even the most stick-in-the-mud customers to
the upgrade table every couple years.

What we need to fix this is to find a joe-the-anal guy who works at a
Really Large customer that spends 100 million a year on MS products
who is irritated with this lack of RDNSS to call up MS and tell them
"your going to add in RDNSS to the next Windows version or else"

Then 3 years from now both will support it and this discussion will
be a no-op.

I don't see the situation as hopeless unless people just take the
attitude that there's nothing we can do.  And no, I'm not advocating
that people deliberately advertise broken IPv6 nameservers.

> But we could discuss whether it currently makes sense to support DNS
> queries over IPv6 at all. Doubling the number of resolver addresses is
> not necessarily good.  I believe most consider 2 addresses optimal in
> IPv4 only deployments.  Resolver libraries are notorioulsy bad at
> fail-over, so there isn't really any point in more that that.  And you'd
> better make both addresses always be up and responding. Adding 2 more
> addresses is only adding trouble.  So why do we do that?  To be nice to
> all the IPv6 only devices out there?  Right...
>

Good point, an even better point is why are the CPE's even in the DNS
chain at all?   These are devices that are mass-produced with a
CPU that is barely fast enough to work and just barely enough ram
for their own NAT/routing processes, running a caching nameserver on
them seems like asking for a smackdown.

>
> Bjørn