1 |
On Tue, 27 Feb 2007 09:11:33 -0800 |
2 |
Grant <emailgrant@×××××.com> wrote: |
3 |
|
4 |
> > > > > > Anyway, a closed port remains closed whether a firewall is |
5 |
> > > > > > running, or not. |
6 |
> > > > > |
7 |
> > > > > I thought the firewall specified which ports to open/close. |
8 |
> > > > |
9 |
> > > > Not quite, but we might be running into terminology here. |
10 |
> > > > |
11 |
> > > > The app that is listening a port opens the port. This has |
12 |
> > > > nothing to do with the firewall. The firewall is simply an |
13 |
> > > > extra level of checks applied before the packet is allowed |
14 |
> > > > thorugh the firewall to be received by the kernel, in the same |
15 |
> > > > way that a bouncer allows or disallows the public to enter a |
16 |
> > > > club. If the bouncer is off sick, the public gets to walk |
17 |
> > > > through the door up to reception, assuming the club is open for |
18 |
> > > > business. |
19 |
> > > > |
20 |
> > > > What Mick was referring to is that if a service is running, it's |
21 |
> > > > still going to listen on it's port whether iptables is running |
22 |
> > > > or not. So, in the absense of iptables (i.e. your bouncer is off |
23 |
> > > > sick), you hopefully have a decent password strategy in use by |
24 |
> > > > whatever is actually listening on the box. |
25 |
> > > |
26 |
> > > So as far as incoming connections are concerned, if there are no |
27 |
> > > listening applications, there is no need for a firewall? |
28 |
> > |
29 |
> > Technically yes. In the real world, it depends. The theory will |
30 |
> > work if and only if you can absolutely guarantee that no listening |
31 |
> > service will ever be running behind that firewall, and that this |
32 |
> > will always be true from here on out till the end of time |
33 |
> > regardless of who has access to the machine. |
34 |
> > |
35 |
> > That's a tall order, and leaves human nature out of it. You might |
36 |
> > install a listening app and leave it running in error without |
37 |
> > realising the impact of not having a firewall. Someone else might |
38 |
> > do the same. |
39 |
> > |
40 |
> > Ubuntu takes the approach you just asked about and it mostly works |
41 |
> > well, especially for notebooks on a LAN behind a NATing gateway. If |
42 |
> > you are running a network with valuable private information on it, |
43 |
> > you might well prefer a belts and braces approach of having a |
44 |
> > mostly-closed firewall as well. |
45 |
> > |
46 |
> > As always, the best solution will vary according to what *you* need |
47 |
> |
48 |
> On more question, is default the right runlevel in which to run |
49 |
> shorewall? It looks like it's one of the last services to start that |
50 |
> way. |
51 |
> |
52 |
> - Grant |
53 |
You could probably run it sooner, but it might try to bring up your |
54 |
network interfaces before starting. But, as long as the interfaces' |
55 |
device nodes exist, whether or not the connection is up shouldn't |
56 |
matter. You could move it to the boot level, probably, if you were |
57 |
worried about security while the computer boots. |
58 |
|
59 |
-- |
60 |
gentoo-user@g.o mailing list |