I think the original post was ,of course not intended, misleading about
how the owner was using the word "Proxy". A "Proxy" usually is used in
the meaning " someone of something, standing IN for another party that
is not directly addressable, <NAT> would be one proxy. I believe, I was
in the mind frame that the post was questioning is it, would be feasible
to have the IPF-Bridge firewall make connections for him on his behalf
and not merely just pass ,unmodified, the original packet if it matched
a rule in the FILTER.  

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]] On Behalf Of Jefferson Ogata
Sent: Wednesday, June 12, 2002 6:42 PM
To: IP Filter
Subject: Re: FTP Proxy with IPF 3.4.28

Clayton Fiske wrote:
> On Wed, Jun 12, 2002 at 02:45:30AM -0400, Jefferson Ogata wrote:
> 
>>The FTP proxy doesn't theoretically require an IP address. It just
sits inline 
>>and adds rules as it observes PORT commands. It may not work, but
there's no 
>>reason it couldn't work theoretically in a bridge configuration.
>>
>>There's no real reason NAT couldn't serve a limited function in a
bridge 
>>configuration as well. Just because an address gets translated doesn't
mean 
>>that the resulting address must reside on the firewall. It just needs
to have 
>>arp in place so it gets routed back. Again, I'm talking theory here,
not practice.
> 
> 
> I think this is all just a matter of semantics. In fact, NAT and
> proxies do require an IP address. They just don't require it to be
> assigned to an interface on the firewall. But if you're going to
> add an arp entry pointing it at the firewall anyway, then for the
> purposes of NAT and proxy the firewall does have an IP. It just
> doesn't respond directly to it.
> 
> But this is all just splitting hairs.

I don't think it's splitting hairs. NAT is a completely different
function 
from what the IP Filter FTP proxy does when it is not asked to perform
address 
translation. In non-translating mode, the FTP proxy is simply a filter
rule 
engine.

As for arp, I did not say that the arp entry should point at the
firewall; I 
said only that arp needs to be in place. With the firewall in bridging
mode, 
this means that the arp for the source address must point to a host on
the 
other side of the bridge.

Even if, however, you do point arps at the firewall's interfaces, that
does 
not mean the firewall has an IP. If the firewall has an IP, that means
you can 
address services bound by the firewall's bind() call using that IP, and
arp by 
itself doesn't accomplish this at all. It is perhaps, semantics, but it
is 
definitely not splitting hairs.

-- 
Jefferson Ogata <[EMAIL PROTECTED]>
NOAA Computer Incident Response Team (N-CIRT) <[EMAIL PROTECTED]>



Reply via email to