Real-time chat systems, such as UNIX talk, IRC, ICQ, and various online service chatrooms, have been in existence since the earliest days of the Internet. The first chat occurred in 1972, between Stanford and BBN. 
Chat systems come in two models from the user perspective: real-time chat and instant messaging. In a real-time chat, messages from many users appear one after another on the same screen, as soon as they type them. An electronic chatroom is analogous to a busy cafeteria, where several conversations can be going on at once, except in an electronic chatroom it is possible to "listen to" or even "record" every one of them.
In instant messaging, a user's message appears on the recipient(s) screens as soon as it is sent. The recipients can respond via another instant message. Instant messaging is more like extremely fast e-mail, and usually involves only two people at a time.
From a technological perspective there are also two models of chat systems: client-server and peer-to-peer. In client-server architectures, such as IRC or AOL instant messenger, a user's message is sent to a central server (or network of servers), from which it is redistributed to other users connected to that server (or network). In peer-to-peer systems, such as ICQ or NetMeeting, users connect directly to one another to communicate after obtaining enough information to do so from a central server.
ICQ is a peer-to-peer chat system whose popularity has exploded in the past year.  ICQ clients connect to a server to announce their presence online. The server then alerts the client when any member of a user-defined list comes online or goes offline. Clients then connect directly to one another to chat or send instant messages.
Governance on such systems in general is on an individual level, much like governance of e-mail. In other words, individuals can choose the people with whom they interact. In theory, ICQ users can "hide" from people and prevent people from determining if they are online. However, the ICQ protocols are not security-conscious and have been widely reverse engineered.     The reality is that an ICQ user can impersonate other users, gain information about them without their knowledge or consent, and bypass any attempt by another user to ignore them. ICQ communities are thus effectively anarchic.
America Online (AOL) employs a client-server model for chatting and instant messaging. Users log in to the service, and are able to join chat rooms or send messages to other users through the server.
The technological possibilities for governance are greater for AOL than for ICQ, because of its purely client-server nature. Like ICQ, users can ignore or make themselves invisible to other users. Unlike ICQ, these measures are not easily circumvented. Users with grievances can report "Terms of Service (TOS) violations" to the company, which can take appropriate action, including disabling the offending user's account.
Recently America Online has begun to offer a service called AOL Instant Messenger, which allows non-AOLers to exchange instant messages with AOL members. Misbehavior can be regulated in a similar manner, however, AOL Instant Messenger also features a "warn" button, which warns a user whose behavior is offensive. Too many warnings and IM ceases to function for that user. 
Created in 1988 by Jarkko Oikarinen at the University of Oulu in Finland , Internet Relay Chat (IRC) is another client-server chat system. Each IRC server, usually a part of a larger network of servers, carries a set of channels. Users can be present on several channels at once. Messages sent to a channel are rebroadcast immediately to all users viewing that channel. Messages sent by a user on one server are relayed to the other servers in the network, so users can connect to a channel on any IRC server in an IRC network and see messages from people connected to any other IRC server in that network.
IRCops can kick users off certain channels (chat rooms), kick them off the server, ban them from channels, ban them from the server, or even ban their entire site from the server. Operators of existing servers will not allow a new server to connect unless they know and trust its operator; this practice is designed to ensure the quality of IRCops.   Governance of IRC is covered in more detail in our case studies.