Internet Relay Chat Operator Etiquette Guide January, 1991 Welcome! You've either been selected to be an IRC Operator or you have set up your server and thus have taken on the dual task of IRC Server Administrator and IRC Operator. Your future days will be filled with hours of fun chatting on IRC, and then wondering why everyone you talked to went away, because the links had apparently broken. Linking: ======== You will be assigned links from the IRC Routing Coordinators. Please use these links and these links ONLY. The links have been designed to maximize efficiency and make delays in chatting minimal. You will be given two links, one to each regional backbone site. Connect to the primary site first and then to the secondary site. You should not need to connect to any other sites. You will be informed if this policy changes. Kills and Walls: =============== /kill and /wall are special operator commands. You should use them with care, and only if absolutely needed. The format are as follows: /kill USERNAME comment. comment can be a phrase of almost any length (within reason) and should be used for specifying the reason of the kill. example: /kill Trillian She's a Ghost IRC Ghosts are created after a net split has occured and the net has yet to relink. /wall PHRASE. This is used for an emergency command like the net is about to split into little pieces, and everyone should reconfigure their links as soon as possible. You will see a WALL when it happens, an operators nickname will appear with # signs around it. #Trillian# Server bucsd.bu.edu coming down for upgrade. Prepare to reconfigure links. /wallops PHRASE This is used to talk to ALL operators at once. It is not often warranted, but is useful. Often, when there is an important IRC situation that requires all the operators attention, /wallops is used. The form for wallops is a nickname with ! signs around it. !Trillian! Australia should leaf off of eris, not carry all the US traffic. /TRACE command /TRACE is useful to know what servers are connected to what. Sometimes /trace can be confusing, especially you are using it for the first time. Here is an example of a trace from bucsd.bu.edu to betwixt.cs.caltech.edu. /TRACE betwixt.cs.caltech.edu IRC log started Mon Aug 26 17:04 *** Link eff.org<2.6.1a> ==> h.ece.uiuc.edu -bucsd.bu.edu- *** Link bucsd.bu.edu<2.6.1a> ==> h.ece.uiuc.edu -h.ece.uiuc.edu- *** Serv Class[3] h.ece.uiuc.edu ==> ucsu.colorado.EDU -h.ece.uiuc.edu- *** Serv Class[2] h.ece.uiuc.edu ==> bucsd.bu.edu -h.ece.uiuc.edu- *** Serv Class[10] h.ece.uiuc.edu ==> silver.ucs.indiana.edu -h.ece.uiuc.edu- *** Serv Class[11] h.ece.uiuc.edu ==> monitor.ece.uiuc.edu[h.ece.uiuc.edu] -h.ece.uiuc.edu- *** User Class[0] h.ece.uiuc.edu ==> Razorbone[uxa.cso.uiuc.edu] -h.ece.uiuc.edu- *** Serv Class[10] h.ece.uiuc.edu ==> MHD54.MOORHEAD.MSUS.EDU[134.29.97.1] -h.ece.uiuc.edu- *** Serv Class[4] h.ece.uiuc.edu ==> *.umich.edu[mingin.engin.umich.edu] -h.ece.uiuc.edu- *** User Class[0] h.ece.uiuc.edu ==> BooServ[sunc4.cs.uiuc.edu] -h.ece.uiuc.edu- *** Serv Class[10] h.ece.uiuc.edu ==> nic.stolaf.edu -h.ece.uiuc.edu- *** Oper Class[0] h.ece.uiuc.edu ==> SodaPop[h.ece.uiuc.edu] -h.ece.uiuc.edu- *** Serv Class[4] h.ece.uiuc.edu ==> oddjob.uchicago.edu -h.ece.uiuc.edu- *** User Class[0] h.ece.uiuc.edu ==> Deviant[isca01.isca.uiowa.edu] -h.ece.uiuc.edu- *** Serv Class[10] h.ece.uiuc.edu ==> *.uc.edu[hilbert.che.uc.edu] -h.ece.uiuc.edu- *** Class 0 Entries linked: 4 -h.ece.uiuc.edu- *** Class 11 Entries linked: 1 -h.ece.uiuc.edu- *** Class 10 Entries linked: 4 -h.ece.uiuc.edu- *** Class 4 Entries linked: 2 -h.ece.uiuc.edu- *** Class 3 Entries linked: 1 -h.ece.uiuc.edu- *** Class 2 Entries linked: 1 This shows that from eff.org (running version 2.6.1a to h.ece.uiuc.edu, routing goes through bucsd.bu.edu before reaching h.ece.uiuc.edu. "h" is connected to ucsu.colorado.edu, silver.ucs.indiana.edu, monitor.ece.uiuc.edu (which resolves to [h.ece.uiuc.edu], hence the square brackets), mhd54.moorhead.msus.edu, *.umich.edu (which resolves to mingin.engin.umich.edu), nic.stolaf.edu, oddjob.uchicago.edu, *.uc.edu (which resolves to hilbert.che.uc.edu), and bucsd.bu.edu, which is known as its "uplink". It is quite normal for a server to have several "uplinks" if it is busy. An uplink is the link towards the top of the tree which carries alot of traffic. "h" also has several users on it, RazorBone, BooServ, and Deviant. SodaPop is also an active operator on h.ece.uiuc.edu. You can tell each host from what each person is logged into by looking just past their nickname. For example, Deviant is logged in from isca01.isca.uiowa.edu. /SQUIT server {comment} /squit isolates a specified link from the next closest uplink server. This is usually used in conjunction with CONNECT (explained later) to reroute traffic. This will be described in detail in the section "routing", preceding CONNECT. SQUIT can be used in one of two ways. It can be used on a local server, which would cause the server name specified to be unlinked relative to your path; or you can send a message to another server instructing it to issue an SQUIT to that server, in which case the break will be relative to the remote server. Usage (and examples): /squit E If the network looks like this initially (and you are on server A) A <---> B <---> C <---> D ^ | v G <---> E <---> F <---> ... (rest of the net) Then after issuing the previous /squit the network would look like this: A <---> B <---> C <---> D G <---> E <---> F <---> ... /squit E {comment} It usually helps to give a reason why you are sending a SQUIT for a server. This can be accomplished by sending the command "/squit server This link is making the US route through Finland". The SQUIT will then be sent out, and the server sending the squit will WALLOP sending the comment so all operators can see it. /CONNECT server {portnum server2} /connect is used to establish a link between two servers. These connections must be authorized by each server's ircd.conf file, but any operator can issue a CONNECT between authorized servers. This command is most often used in conjunction with SQUIT to reroute traffic. If only one argument is given, this command causes the server you are on to attempt to connect to the server specified. For example, "/connect B" (in the previous example) would cause your server (A) to connect to B. Suppose you wanted to reconnect server F to server E? You cannot contact server F since it is no longer part of your network. However, you can tell server E to connect to it. A remote CONNECT can be issued to server E. Examples (assume you are on server A): /connect B If the network initially looks like this: A B <---> ... (rest of network) Then afterwards (if the connection succeeds) the network will look like this: A <---> B <---> ... In the example where you wanted to reconnect server E to F, the following syntax would be appropriate (note: we are assuming that F's irc socket port is 6667, which is the default) /connect F 6667 E If the network initially looks like this: A <---> B <---> C <---> D ^ | v G <---> E F <---> ... Then after your CONNECT request the network topology will look like this: A <---> B <---> C <---> D ^ | v G <---> E <---> F <---> ... Be careful when connecting servers that you know which command to use! If you simply issued "/connect F" from your server, the network would look like this: ... <---> F <---> A <---> B <---> C <---> D ^ | v G <---> E which for various reasons (discussed below) might be very undesirable. Routing ======= When and how should you do rerouting? This depends on where your server is topologically located and whether you route traffic. If you are a leaf node (i.e. only connect to one server at a time) then chances are you won't need to do any routing at all. Your ircd.conf file should be written to connect to the best possible servers first before trying alternates. At the most, you may decide to squit an alternate server and connect to your primary if/when it goes back up. This only involves local squits, however. If you are operating a backbone site, you may find yourself rerouting things quite often. If the EFnet servers (see the file doc/networking) badger.ugcs.caltech.edu (Pasadena, CA), irc.mit.edu (Boston, MA), minnie.cc.utexas.edu (Austin, TX) and ucsu.colorado.edu (Boulder, CO) were routing traffic in the following way: ... <---> minnie <---> badger <---> bucsd <---> ucsu <---> ... It would make sense to either squit ucsu and reconnect it to minnie, or disconnect minnie from badger and connect to ucsu, because topologically (and geographically) ucsu and minnie are rather close. There are occasions when US traffic for some reasons winds up being routed through Australia. This is another case where traffic should definitely be rerouted. However, there are sometimes occasions when routing is going through "backdoor" methods. If you see something totally outrageous (like the east coast and the west coast being connected by eff.org) please /WALLOPS and ask before you send any squits, because chances are, it's like that for a reason. Of course, any operator can remotely squit or connect servers, so if you see a problem and you're sure you know how to fix it, it's a good idea to do so. If the operator of a server which is is being routed poorly is online, it's probably best to contact him/her first, though. Chances are that hub operators will be more familiar with the general topology of the network and which servers connect to which (which is why most of the manual routing is left to them), so if you have any problems, talk to the other operators via /wallops. That's what it's there for! Also, be aware that servers will notify all the operators online of remote SQUITs and CONNECTs via WALLOPS. Please let us know if there should be any additions to this guide. Again, this is not MANDATORY, this is just a GUIDE. Please conduct yourself as an IRC Operator would...you are looked upon for assistance, both emotional and mental. Helen Rose Christopher Davis Noah Friedman January, 1991