Re: Reducing nick-collision takeovers

[Christophe Kalt <kalt@xxxxxxxxxxx>]

On Sep 04, Kaspar Landsberg wrote:
| Since takeovers are most effective when they are done by nick-collisions,
| the following idea might well reduce the number of takeovers by reducing the
| number of people killed by a nick collision:

how often do this occur? (both ND and CD have timed out, it
seems)

are splits commonly long enough that ND and CD both time
out?

on small networks, splits shouldn't be long (and the default
delay is fine),

on larger networks, splits can be longer.  It's easy to
deal wit this administratively.  track the servers commonly
used by abusers and deal with them (raising the delay is one
option).
	
| When a client (user) registers (right after connecting to the server), one
| could add for each new user a random number (between 0 and n) to its
| structure. Then, when a nick collision occurs, the two involved servers
| check the random number of 'the other' user. If, for example, it's higher 
| than the number of the local user, the latter gets killed. If it's lower,
| the local user doesn't get killed. If both random numbers are equal, both
| users get killed.

Do you realize that this is TS under disguise?

| PPS I do still think, 'forced nick change' is so far the best approach.

fnc requires major change(s) to the protocol, and isn't
possible to easily implement gradually.
of course it's possible, but on top of that, fnc is likely
to be disruptive.