On Mon, Jul 16, 2001 at 03:08:51PM -0400, Kev wrote:
> *** Net junction: Test0.undernet.org Test1.undernet.org
> *** Nick collision on KevveK (freebsd.testnet.org 994978070 <- Test1.undernet
> .org 995309412 (Different user@host))
> *** Completed net.burst from Test1.undernet.org.
> *** Test1.undernet.org acknowledged end of net.burst.
> !freebsd.testnet.org! Protocol Violation from Test1.undernet.org: Too few
> parameters for BURST
>
> I finally understand this. On Test1.undernet.org, KevveK joined a channel
> and set no modes. On the Test0.undernet.org side, however, KevveK already
> existed. Test0.undernet.org received the NICK for a KevveK and noted a
> nick collision. Later, when the BURST for the channel came, it looked
> something like: "BF B #test 995069190 BFAAA:o". Test0.undernet.org said,
> "There's no BFAAA", but propagated the BURST anyway. We decided to propagate
> the empty BURST on the off chance that everyone in the first BURST message
> got collided, but the BURST happened to be split across multiple messages--
> since the channel modes and timestamp are only sent with the first BURST
> message. However, in the case of a channel with no modes, that empty BURST
> triggers the protocol violation notices that many have seen and pointed out.
>
> Question is, what should we do about this? My vote is to simply back off on
> the required parameter count for m_burst(). Anyone have other opinions?
Sounds good by me :)