At 14:42 -0600 2/4/99, Fuzzy Fox wrote:
>Clifford Hammerschmidt <[EMAIL PROTECTED]> wrote:
>>
>> I though that since it was build into the kernel it also modified the
>> code for creating sockets so user sockets where from 1024-61000,
>> rather than 1024-64k. (In theory you should be able to place MASQ in
>> any 4k range and have it configure that regular socket calls to avoid
>> that space right?)
>
>Maybe. I haven't looked deeply enough to find out if that's the case.
>At any rate, it doesn't matter, because your masq box is not the only
>computer that chooses socket port numbers.
In fact the only choices that count are those made on the masq box, other
boxes don't really matter. If another box attemts to 'do something' to an
arbitrary port on the masq box, some program on the masq box must first
have chosen that port in order for anything predictable/useful to happen.
Except for oddball utilities like traceroute, which take pot-luck on some
port number (33434 in traceroute's case). Traceroute actually hopes that no
program on the masq box has chosen that port to communicate with, so it
gets the expected ICMP response.
Different applications running on the masq box might choose specific ports
>1023, e.g. VNC uses ports in the 580x and 590x ranges (in addition to the
X protocol's 600x).
If the kernel truly disallowed programs on the masq box from obtaining port
numbers in the masq range, the only conflicts which might result would be
with applications running on the masq box that specifically request ports
in that range, which would fail to work. The only problems from other boxes
would be programs that expected that program supposedly running on the masq
box to have succeeded in getting the port. Or oddball applications like
traceroute.
--
Mike Casteel
[EMAIL PROTECTED] Seattle, WA
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
For daily digest info, email [EMAIL PROTECTED]
