Great Circle Associates Firewalls
(February 1995)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Some questions
From: jdb @ ecofin . ch (John B*hrer)
Date: Thu, 2 Feb 1995 15:05:00 -0100
To: firewalls-digest @ GreatCircle . COM (FireWalls Digest)

Date	2.2.95
Subject	Some questions
>From	John B*hrer
To	FireWalls Digest

Some questions                                                           
(Internet)
   

1) I've seen this diagram proposed in recent postings:

   Internet -- Router1 -- Firewall -- Router2 -- Internal Network

What's the purpose of Router2 ?  That is, what extra security does it
provide? Can't I accomplish the same thing with proper configurations of
Router1 and the Firewall itself?  Assuming that the Router -- Firewall
connections are in fact subnets (and not some sort of serial link), I would
call the first subnet (Router1 -- FireWall) a DMZ, for low-security hosts
visible to the outside world. Functionally speaking, what would you call the
subnet between the Firewall and Router2 ?


2) I see the point of this configuration:

   Internet -- Router1 -- Firewall -- Internal Network

but recently someone proposed this:

   Internet -- Firewall -- Router1 -- Internal Network

Although this makes it easier to restrict internal users from going outside,
does this offer any more security against the outside world coming in ?  I
can see that the latter supports more logging than the first scenario, ie,
when the router disallows an inbound packet, this generally isn't logged,
but a firewall computer can do it.


3) FWTK vs. Socks: I'm wondering about this too.  Any objections if I run
both systems on the Firewall, dividing my services accordingly ?  For
example, I like the convenience of Socks FTP, but I suspect that the TIS
http-gw can offer more security for WWW.


4) Subnets: Class-A's were gone long ago, and a mere mortal can't get a
Class-B, so we "grass roots" administrators must put up with a single
Class-C address for our sites. Of course I need sub-networks, but the policy
of "forbidding all zeros / all ones" in a subnet address is just too
restrictive. I don't care what the RFC says, I'm not going to throw away a
good chunk of my address range just to install subnets!  (eg: one-bit subnet
= throw away ALL usable host addresses, 2-bits = throw away half, 3-bits =
throw away 25%, 4-bits = limited to 14 machines per subnet, no way!)

Given that I'm ignoring this discriminatory restriction and doing it anyway
(I don't support broadcast-to-all-nets or "this-network" references), do you
see any security holes as a result? Does it matter if my DMZ subnet is the
"illegal" upper network with a mask of 255.255.255.192 ? I'm using Cisco
routers if that makes any difference. Who uses these reserved subnet bands
anyway?


  John Buehrer      jdb @
 ecofin .
 ch
  Ecofin, AG        phone: +411 / 201 68 33
  Lavaterstr. 45    fax:   +411 / 202 89 47
  CH-8027  Zurich
  Switzerland




Follow-Ups:
Indexed By Date Previous: unsuscribe firewalls
From: Kalpesh Unadkat <kalpesh @ ciesin . org>
Next: Re: tweaking PC setups for TCP/IP
From: morgan @ engr . uky . edu (Wes Morgan)
Indexed By Thread Previous: Compiling S/Key on Solaris hosts
From: Mohamed Ellozy <ellozy @ netman-mel . dfci . harvard . edu>
Next: Re: Some questions
From: mcr @ milkyway . com (Michael Richardson)

Google
 
Search Internet Search www.greatcircle.com