DaZZa wrote:
[EMAIL PROTECTED]">
On Tue, 5 Mar 2002, Andy Eager wrote:

Certainly pretty good as far as a basic explanation goes, problem is
that masquerading is not yet up to the level of ipchains and thats what
most people want. (One IP address, masqueraded to many machines for use
with ftp, realaudio etc). I still reckon that ipchains with a 2.2
kernel is still the simplest and most generally accepted way to do
firewalling if you want particular services masqueraded.

I'm interested to know your reasoning here.
OK, I'm prepared to be knocked down in flames here (only if your gentle), but recently I paid a reasonably well known Linux consultant to advise me on a job I'm doing for a paying customer to install a firewall (that has to do masquerading of all well known services).  His/her sugestion was to stick with ipchains for the time being (preferably under 2.2 kernel because the helper modules were already written and known to work)  After some of my own research, I came up with the following from the man (rusty) himself:

                         494: <sect2>Altering the Destination of Locally-Generated Connections
                         495:
1.15          rusty      496: <p>The NAT code allows you to insert DNAT rules in the OUTPUT chain,
                         497: but this is not fully supported in 2.4 (it can be, but it requires a
                         498: new configuration option, some testing, and a fair bit of coding, so
                         499: unless someone contracts Rusty to write it, I wouldn't expect it
                         500: soon).
                         501:
                         502: <p>The current limitation is that you can only change the destination
1.16          rusty      503: to the local machine (e.g. `j DNAT --to 127.0.0.1'), not to any other
1.15          rusty      504: machine, otherwise the replies won't be translated correctly.
1.1           rusty      505:
                         506: <sect>Special Protocols
                         507:
                         508: <p>Some protocols do not like being NAT'ed.  For each of these
                         509: protocols, two extensions must be written; one for the connection
                         510: tracking of the protocol, and one for the actual NAT.
                         511:
                         512: <p>Inside the netfilter distribution, there are currently modules for
                         513: ftp: ip_conntrack_ftp.o and ip_nat_ftp.o.  If you insmod these into
                         514: your kernel (or you compile them in permanently), then doing any kind
                         515: of NAT on ftp connections should work.  If you don't, then you can
                         516: only use passive ftp, and even that might not work reliably if you're
                         517: doing more than simple Source NAT.

[EMAIL PROTECTED]">


What, exactly, doesn't work under iptables?

I have a 2.4 kernel running iptables, and it seems to do everything fine -
telnet, ssh, ftp, ICQ, irc, real audio, http, https - I haven't found
anything yet that _doesn't_ work.
Are you masquerading all of these (both passive & active ftp?).  I would be keen to hear that there _are_ conntrack modules for the other protocols you mentioned.  

I am more than happy to learn, so please don't hammer me with lots of hurtful flames if I've missunderstood something here.  
(I'm basically a very sensitive person on the inside!! make love not war!!)
[EMAIL PROTECTED]">


So, what am I missing?

DaZZa




Reply via email to