Default security functions on an IPv6 CPE

Ted Mittelstaedt tedm at
Thu May 12 12:53:24 CEST 2011

On 5/12/2011 3:32 AM, Rémi Després wrote:
> Le 12 mai 2011 à 10:54, Mikael Abrahamsson a écrit :
>> On Thu, 12 May 2011, Ted Mittelstaedt wrote:
>> We had an argument here at the office regarding firewall in CPE,
>> and my opinion was that since UNPNv6 isn't really supported
>> anywhere, IPv6 with deny new sessions might even be worse in
>> practice than NAT44 with UPNP. It seems Windows 7 does support
>> UPNPv6, but I have yet to locate information regarding support in
>> shipping software for IPv6 enabled CPEs.
>> So I would like to withdraw my recommendation before about
>> low-ports closed but high-ports open. I now believe that it's
>> better to have a permit ESTABLISHED firewall as default cpe
>> behaviour, and let the clients request policy changes from the
>> gateway if they want other behaviour.
>> We have a social contract with users to have the CPE act the same
>> way it did with NAT, because that's all users know.
> May I challenge this? Ordinary users need plug-and-play and backward
> compatibility, nothing more.They don't ask for NAT compatibility for
> the simple reason that they don't know what a NAT is. (User's of Free
> haven't asked for anything like it.)
> Now, an ordinary customer whose laptop works at home may not work the
> same in his friend's home if this friend has a FW on by default.
> Having to debug such a problem without a site manager in either site
> is IMHO to be avoided.

I don't know how it is in your neck of the woods but here, the
end users call the ISP first.  Even if the problem isn't the ISP

I estimate that 80% of our phone support time is wasted on explaining
to the customer that they have a virus, or their hard drive is
screwed, or whatever.

Back in 2002 I inserted a bit of doggerl code into our mailserver that
changed all instances of a period in any attachment filename to a tilde. 
  That is all it did.

This increased our e-mail support call time about 20% for about a month 
as people found that when they went to save an attachment file that
they couldn't open it, they had to rename it.

It decreased our virus-related support call time by about 60%.

Our typical virus-related support call lasted about 20 minutes,
half of it pretending to sympathize with the customer while they
railed against Microsoft for making a cruddy OS because they were
now going to have to pay lots of money to take the machine somewhere
and have it cleaned.

Our typical explain-about-the-tilde-thing call lasted about 5
minutes and usually the customer thanked us for watching out for
them after we got finished educating them about why clicking on
a file named isn't a good idea.

And after the customers learned about the tilde thing they never
called back.

After the customers got their virus-laden system put back together
they always called back to get configured again - and of course we
got to once again pretend to be interested in another 15 minute
tirade against Microsoft.

This is the reality of ISP support.  That is why ISPs aren't going
to ship CPEs that are hanging wide open.  The support calls from
people unfamiliar with the technology are always cheaper than the
support calls from people unfamiliar with the technology who have
just been gunned.


> Regards, RD
>> So for initial deployment the firewall in the CPE should act the
>> same way the state machine in NAT does, but of course have UPNPv6
>> support so as programs get this, they will be able to interact with
>> the FW.
>> I realise this will limit deployment of true e2e applications, but
>> I believe this is unfortunately needed.
>> -- Mikael Abrahamsson    email: swmike at

More information about the ipv6-ops mailing list