Many people had plenty of experience with instant messaging even before things like Google Talk ever appeared, but between that and things like Facebook's chat function the concept of an in-browser instant messenger has become familiar territory. Wave isn't a substitute for other instant messaging apps -- e.g., AIM -- but more like a parallel venue for real-time discussion.
Like instant messaging, wave discussions are logged as they happen, and the other person's typing can register on your own screen in real time. Unlike instant messaging, though, you're not obliged to respond only to the last thing someone posted -- or, rather, you can comment contextually on previous posts without the conversation derailing itself. If you and your friends in the discussion have a habit of jumping around or engaging in several parallel discussions at once, using Wave to hold that kind of talk imposes order that would be next to impossible to find through a typical chat client.
...But What Is It, Really?
When Wave first premiered, it was widely rumored that it would replace or eclipse any number of other, existing systems and services. E-mail, mainly: Wave has a high degree of built-in security, while e-mail is natively about as secure as sending a postcard written in pencil.
The more people were able to work with it, however, the clearer it became that Wave wasn't intended as a replacement for many things -- and now it's clear that it might not even be intended as an adjunct to them, either. Instead, it's entirely possible that Wave is being used as one of two things.
The first is the "concept car" analogy I mentioned earlier: it's a demonstration of a whole group of different Web 2.0 (and possibly Web 3.0) technologies that could be broken out on their own and put to use in any number of contexts.
The second is a little trickier: Wave is an extended experiment in application interaction -- a way to take many common user interface metaphors (e-mail, discussion groups, IMs, etc.) and re-implement them in new ways. Most of us are so familiar with the concept of e-mail that any thinking about the way it's put together tends to stop right there: there's an inbox, a spam trap, a list of unread messages, etc. Wave's ingenuity is in taking the outward metaphors of many things we take for granted and combining functionality among things that, on first glance, might not seem to play well with each other.
A programmer friend of mine described Wave as "a research project in human-computer interaction." It makes sense: by creating something a great many people will want to try out in an enthusiastic if also provisional way, Google can figure out which parts of the protocol -- both on the backend and in the implementation -- are worth developing, and which parts are best left as add-ons by third parties or discarded entirely. And Google's long made a name for itself as a company that creates things that are experimental by their very nature, with their years-long beta cycles.
Because Wave is so amorphous, many things are missing, and many of those omissions are almost certainly by design. One is a way to migrate to Wave -- for instance, a tool that would let you take your existing e-mail store and convert it into a set of Wave conversations. No such thing exists right now. Not just because no one's written it, but because Wave itself is a moving target, and so migrating to it would be pointless. The protocol could be nothing like what it is now by the time people other than Google start using it. (In theory one could build Wave servers that run in parallel to one's existing e-mail system, create gateways between the two, and then incrementally migrate the functionality of the latter into the former -- but again, why do that when you don't know what you're really migrating to in the first place?)
Another and far bigger issue: Right now, the only version of Wave is Google's Wave. If Wave is meant to be an open protocol that can be implemented by any number of people, either on the client or server side, it'll have to exist in multiple independent implementations before it can be considered any kind of protocol or platform to use in a production sense.
The last word on Wave for now would seem to be that it's aptly named. It's a moving target, and whatever its final incarnation -- if there is one -- it's likely to only resemble what we have now in the most distant way.
For Further Reading
Is Wave A 'Concept Car' For Google?
Enterprise 2.0: Google Wave, A Solution Seeking A Problem?
Google Buys AppJet To Power Wave