On 04/16/2013 03:50 PM, lux-integ wrote:
> On Tuesday 16 April 2013 19:57:24 Bruce Dubbs wrote:
>> Whether to use ipv4 or ipv6 really depends on the ISP.  Generally an
>> office only needs one address and then uses NAT (network address
>> translation) internally.  NAT multiplexes a single address into many.
>> Any office can use private IP addresses internally (192.168.x.x,
>> 10.x.x.x, 172.16,x,x) and one ipv4 (or ipv6) externally.
>
> the problem is  some useful applications such as IPsec and VoIP  do not like
> NAT .

Who says? Only poor implementations don't deal well with it, but single 
NAT should work fine for almost everything now days.

> Also one NAT-router behind another and NAT becomes problematic.
> AND  one has to resort to all sorts  of convolutions such as stun-servers,
> nat-traversal, port-forwarding etc  to attempt  to use these.
>

This is referred to as double NAT and is likely to be really supported 
or supportable in very few scenarios. Don't double NAT!

> here is a far better ( than the one I posted previously ) utube video on ipv6
> http://www.youtube.com/watch?v=9UHZ7eX_f30
>

Many of the people who were conscious of networking in the early '90s 
still consider NAT a poor hack at best. The security provided by NAT is 
only a side effect of having broken everything. Double NAT is broken in 
the same way that NAT is, only doubly, because IPv4 in its inception was 
not intended to be used as it is today. Though I've not inadvertently 
dealt with double-NAT in several years, it was my experience that it 
never worked exactly as intended (at least not with the cheaper devices 
that I could afford at the time).

With that said, NAT has made us quite lazy WRT security. It was quite a 
shock at fist, but my traffic rules increased 4 fold on the v6 side 
compared to my v4 rules when I handed my previously internal only home 
boxes routable v6 addresses (my take was that I had better do it on my 
overly complex home network before tackling it in the field). Now, I 
must state that everything _is_ dual stack, and was from the very 
beginning, because I can only actually use v6 at a handful of places. 
Auto-IP works flawlessly with Linux (and has for at least the past few 
years)-- IOW, no need for an IPv6 DHCP server as long as your router 
supports router advertisement, but you must still provide static DNS 
entries manually if you should want to address your DNS servers via v6 
(your v4 DNS servers will happily return AAAA addresses if requested so 
just handle it via DHCP). Windows and Android devices also work 
flawlessly (preferring v6 to v4 when possible), no idea about OS X and 
iOS (not to be confused with IOS which has worked just fine with v6 for 
several years), though I suspect they too should be fine at this point 
in time.

If you do intend to implement v6, the big thing is that you must be 
conscious of the fact that you are exposing your entire internal network 
to the live internet (if that term still means anything). Don't be 
afraid to be overly cautious. Basically begin by denying everything in 
and out, then add to the top of the stack and work from least traffic to 
most traffic, then most restrictive to least restrictive as you go up 
the chain, being as explicit with your traffic rules as is possible. For 
instance, IP/proto:port in->out goes at the very top, followed by ICMP 
and ICMPv6 rules in the middle, then IP/proto:port out->in near the 
bottom, then deny everything else at the very bottom. You can always 
insert rules to allow more traffic, but if you leave a gaping hole from 
the very beginning...well, you are likely to have a gaping hole when you 
think you are finished. You will also likely need to run dual stack for 
quite some time because v6 adoption is moving very slowly.

My advice, having done it myself: Don't do it yet! At least, not until 
you have a real world need to do so. Unless you are doing it for 
training, or have a real need that cannot be addressed with existing v4, 
is just a headache for very little benefit at the moment.

HTH

--DJ

-- 
http://linuxfromscratch.org/mailman/listinfo/blfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to