On Thu, Apr 11, 2002 at 10:00:52AM +0800, ian perez wrote:
> 
> how bout this site  http://www.linuxguruz.org/iptables/

Also very useful :-)

actually, looking at it a again I come to the conclusion that there are as
many ways of composing a packet filter ruleset as there are people doing it,
so maybe there are no generic ways for filtering...

Or maybe I mean something different, like tried and true ways to allow
specific services to be accessible using an iptables ruleset for the cases I
mentioned (behind the FW, on the FW, on the Internet from behind/on the FW)

Although I understand and know a lot about IP protocols, I'm sure I can
learn more than I know already ;-) But packet filtering is a whole new way
of thinking about protocols. Especially in the case of stateful FW actions.
The ESTABLISHED and RELATED states are helpful and at the same time, they
hide their operation in a gray box (a black box that you can look into if
you dare/are able...). I'm reluctant to try and understand the code that
makes the choices about what is a state, but at the same time, how can I
trust that a packet belonging to a RELATED state is not 'accidentally'
unrelated?

Take ICMP for example, I looked into which types are acceptable to let
through, but I also put in rules that accept RELATED and ESTABLISHED states
in the chain for ICMP packets. Some types that I am willing to accept (3 and
11, unreachable and time-exceeded) don't reach the filter rule, because I
guess they are caught by the RELATED and ESTABLISHED rules that are
ACCEPTed. Of course, I could test it using some logging rules, but I don't
care that much for this particular case. It's just an example.


After writing this, I'm not entirely sure what I'm looking for anymore, but
I do know that I haven't seen a document or tool that provides the
information I'm looking for...

Thanks for the helpful URLs!

Simon

Reply via email to