[1733] in linux-net channel archive
Re: Problem starting gated
daemon@ATHENA.MIT.EDU (Shawn Ruttledge)
Mon Jan 29 22:09:08 1996
From: Shawn Ruttledge <ecloud@goodnet.com>
To: sspoon@clemson.edu
Date: Mon, 29 Jan 1996 13:15:19 -0700 (MST)
Cc: linux-net@vger.rutgers.edu, grenoble@acoustic.com, billg@radix.net
In-Reply-To: <m0tgmaQ-00081WC@chaos.holmes.clemson.edu> from "Lex Spoon" at Jan 29, 96 00:57:05 am
> > Another thing which most docs (such as net2) don't seem to acknowledge is
> > that when you are connected to a typical ISP you don't have any control
> > over their PPP server; ie., I can't just add my second machine positron
> > to _their_ routing table.
>
> Well, looks like you figured it out! You need to get real IP addresses
> from your ISP, assuming they are giving you full connectivity as
> opposed to using slirp. If you just use made up numbers, then the
> outside world has no idea where to route the return packets.
But wouldn't dynamic routing on their end fix that (I mean goodnet's end)?
Did my scenario at the bottom of the last message (about dynamic routing)
seem reasonable? It does bother me that the machine has two IP addresses at
the same time, but in general it would seem there isn't a way around it.
Even if I didn't have ethernet, just had ppp with an "official" assigned
address, and then added ax.25, well, amateur radio addresses are always
44.something, so it forces me to have two TCP/ip addresses. And after all,
that's what gateways are supposed to do - accept packets for various machines
and route them to the machines. It would seem that an ISP's machine has to be
a gateway. But on second thought, the point wasn't that my machine has two
addresses, but that the outside world might not be seeing it that way.
The question remains then: is it true or not that a typical ISP does
dynamic routing and that the result of it is that any addresses which appear
under the assigned one will be seen as new addresses and automatically added
to the routing tables? Or do they enforce subnetting, ie check to see if
a "new" address ought to be under the ISP's subnet before adding it?
OK, so assuming that it's not possible for me to get packets back that
are directed to the ethernet addresses I have assigned my machines, what
do I do? How about proxy arp? As I understand it, proxy arp forces a
mapping of tcp/ip addresses to ethernet addresses. But it is superior to
static routing somehow isn't it? I don't understand the difference really.
But if somehow I could get a packet back from that distant www server that
on the outside is addressed to the address my ISP assigned me for the PPP
session, electron will get it, and then it needs some means of looking inside
the packet to see that the packet really ought to go to positron. Hmmm,
sounds like the ip-ip tunneling they use on the packet networks to create
an internet wormhole from one packet region to another. But that isn't a
generic service, it requires stuff on both ends... ahah, how about this:
electron monitors outgoing packets. It notices that positron originated
something to www.mit.edu let us say. Maybe, it even notices that the packet
was http-looking. So it would seem reasonable that if a packet comes back
from that same address ("same" meaning either the numeric address is the
same, or, the DNS name is the same) it probably is a response and positron
was the one waiting for a response. As long as a user on electron doesn't
try to access the same site at the same time it might work. It sees the
packet coming back, and realizes it's probably for positron, so it re-writes
the numeric address as positron's address and sends it over the ethernet
(assuming packets on ethernet still have tcp addresses in them; if not, then
I guess it doesn't need to re-write the destination numeric address, but
just map to positron's ethernet address). Now, assuming we did that, what
if there was a mistake and it really wasn't for positron? Is positron going
to say "no thank you" or accept the packet anyway and throw it away or try and
decide what to do with it? Either way, we could use a semaphore to keep
electron from accessing the same site at the same time, in which case it
would be guaranteed that an http packet from www.mit.edu could only be
going to one machine (which ever one raised the semaphore). Now if the
webmaster at www.mit.edu tried to telnet to one of my machines at the same
time, that would be a different matter....
Anyhow, for now to browse the web from positron, I just get on electron,
nfs mount positron's hard drive, get back on positron, and start positron's
netscape from electron on a remote x session. Makes for a lot of ethernet
traffic and slow loading but at least netscape thinks it's running on
electron and so I can get out to the net. Of course I should just put
netscape on electron's hard drive to save the nfs traffic...
--
_______ KB7PWD @ KC7Y.AZ.US.NOAM ecloud@goodnet.com
(_ | |_) html: http://www.goodnet.com/~ecloud
__) | | \__________________________________________________________________
* eschew obfuscation * ham radio * ARS * electronic music * electronics *
