Hi,

On 12 April 2016 at 23:58, Russell Coker <[email protected]> wrote:

> On Wed, 13 Apr 2016 11:10:35 AM John Mann via luv-main wrote:
> > The problem is that Anycast 6to4 is an _unmanaged_ tunnel system,
> > dependent upon the generosity and technical ability of strangers (often
> > with no means of identifying or contacting them).
>
> The Internet is entirely dependent on the technical ability of strangers.
> Usually the random unpaid strangers are more skillfull and diligent than
> the
> people employed by ISPs.
>

The people advertising the routes (2002::/16 and 192.88.99.1/32) to
gateways are likely to be ISPs.

Remember that 6to4 packets  6->4 and 4->6 are likely to use different
gateways in each direction, take different paths etc etc.
What happens when some network manager somewhere decides to not allow IP
Protocol 41
or decides to not allow ICMP packet-to-big ?
You can send 6to4 packets but they don't get through, and it isn't obvious
where the problem is.

Quick exercise:
If an IPv6-connected client wanted to send a 6to4 packet to your 6to4
server, they would use a 2002::/16 route.
The only public ones are listed at
     http://bgp.he.net/net/2002::/16
Which of these is closest one to each possible client machine??

When your 6to4-using server wants to send a reply packet, they tunnel it
and send back to  192.88.99.1
The only public ones are listed at
    http://bgp.he.net/ip/192.88.99.1
Which one will your server use?  Where is it?

The nearest one to Monash Uni is in the Netherlands.
---
[nismgr@ns0b ~]$ traceroute 192.88.99.1
traceroute to 192.88.99.1 (192.88.99.1), 30 hops max, 40 byte packets
 1  drc1-gw-v10.net.monash.edu.au (130.194.10.253)  0.644 ms  0.688 ms
 0.712 ms
 2  drc0-gw-t2-4.net.monash.edu.au (130.194.28.150)  0.645 ms  0.683 ms
 0.761 ms
 3  monash2-gw-v526.net.monash.edu.au (130.194.28.106)  7.431 ms  7.565 ms
 7.447 ms
 4  138.44.64.56 (138.44.64.56)  1.293 ms  1.289 ms  1.304 ms
 5  et-7-1-0.pe1.wmlb.vic.aarnet.net.au (113.197.15.26)  2.836 ms  2.870 ms
 2.842 ms
 6  et-1-3-0.pe1.mcqp.nsw.aarnet.net.au (113.197.15.8)  15.557 ms  15.592
ms  15.591 ms
 7  et-0-1-0.pe1.rsby.nsw.aarnet.net.au (113.197.15.3)  16.171 ms  16.172
ms  16.148 ms
 8  113.197.15.159 (113.197.15.159)  16.270 ms  16.551 ms  16.485 ms
 9  et-1-0-0.pe1.wnpa.akl.aarnet.net.au (113.197.15.77)  38.682 ms  38.766
ms  38.746 ms
10  et-0-1-0.199.pe1.tkpa.akl.aarnet.net.au (113.197.15.71)  39.120 ms
 39.101 ms  39.200 ms
11  et-1-2-0.pe1.a.koa.aarnet.net.au (113.197.15.85)  118.302 ms  118.244
ms  118.639 ms
12  et-1-1-0-1100.bdr1.b.lax.aarnet.net.au (113.197.15.87)  163.669 ms
 163.630 ms  163.642 ms
13  207.231.246.7 (207.231.246.7)  317.676 ms  317.665 ms  317.617 ms
14  us-chi.nordu.net (109.105.97.81)  335.650 ms  335.607 ms  320.645 ms
15  us-ash.nordu.net (109.105.97.134)  328.637 ms  326.716 ms  320.526 ms
16  nl-sar.nordu.net (109.105.97.138)  333.084 ms uk-hex.nordu.net
(109.105.97.140)  318.742 ms  320.626 ms
17  nl-sar.nordu.net (109.105.97.124)  328.494 ms
GE-1-0-0.3630.JNR02.Asd001A.surf.net (109.105.98.34)  331.221 ms
nl-sar.nordu.net (109.105.97.124)  326.672 ms
18  V1131.SW14.Amsterdam1.surf.net (145.145.19.170)  331.243 ms  333.669 ms
 330.992 ms
19  192.88.99.1 (192.88.99.1)  330.788 ms !X  330.666 ms !X
V1131.SW14.Amsterdam1.surf.net (145.145.19.170)  325.059 ms
---

Adding 100s of ms of latency won't be good for performance!


> With this sort of thing there are risks of hostile parties running tunnels,
> but all unencrypted data on the Internet is at risk of interception so I
> don't
> think that changes much.
>
> Thanks to 6to4 when I get my home Internode connection working with IPv6 I
> will have every system that matters to me on IPv6.  It will be some time
> before everything works well on IPv6, now would be a good time to get it
> all
> tested.


6to4 connection failure rate used to be ~ 20% but is now ~ 10%
   http://www.potaroo.net/ispcol/2015-11/v6perf.html
   see fig4 -> fig7

Luckily, modern clients are now doing less 6to4
  https://tools.ietf.org/html/rfc6724
---
10.7.  Configuring 6to4 Preference

   By default, NATed IPv4 is preferred over 6to4-relayed connectivity:
---
see graph from previous reference:
    http://www.potaroo.net/ispcol/2015-11/fig13.jpg


-- 
*John Mann*
Network Architect, Infrastructure Automation & Delivery

*Infrastructure Services, eSolutions*
​Monash University
738 Blackburn Road
Clayton VIC 3168
Australia

T: +61 3 9905 4774
M: +61 419 568 470
E: [email protected]
W: monash.edu
_______________________________________________
luv-main mailing list
[email protected]
https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main

Reply via email to