Re: [MEL-NAP] Putting AUIX local domains into squid

    If it is only a matter of determing whether host is local or behind
 the border gateway, one might define some sort of localness radius in
 hops, and then construct and send a single udp traceroute packet to
 the host of interest, then watch for the gateway that returns icmp.

 If your border gateway is always say 4 hops away, you can send a udp
 that will be dropped in 4 hops by exact this gateway. It would be
 fairly easy to determine whether host is behind a slow link by simply
 expecting this gateway to drop foreign probes - anything other is local.

 In ideal case there will be one probe - one icmp, with pretty fast
 responce. Of course, probes may be dropped or blocked by firewalls,
 often changing or very complex environments will be hard to deal with.
 In any case this would be a really dirty hack, although it might solve
 the problem for most users with simple topologies.

> I have the same situation here with the Philippines IX, which is here in
> Manila where all the IPLs terminate.
>
> > Thus, squid could interrogate the host's routing tables directly to
> > determine the localness of a host, working on the basis that if the
> > returned route is not 'default' it should be asked directly, and not
> > through a parent. Running gated on the squid machine would thus manage
> > the routing tables for squid.
> >

> PH-IX
> |
> leased
> line
> |
> router2
> |
> router1---main cache----router0
> | |
> pacific ocean
> | |
> sprint mci
>
> If I were to traceroute to a site, I know that once it crosses router0, it
> is a foreign site. When it crosses router2, it's a local peer on the IX.
> If I wanted to be more specific I would watch out for the routers of our
> peers.

-------------------------------------------------------------------
 Andres Kroonmaa Telefon: 6308 909
 Network administrator
 E-mail: andre@ml.ee Phone: (+372) 6308 909
 Organization: MicroLink Online
 EE0001, Estonia, Tallinn, Sakala 19 Fax: (+372) 6308 901
-------------------------------------------------------------------
Received on Tue Jul 29 2003 - 13:15:40 MDT