Recently, there has been a significant rise in the noise from the media (http://blogs.zdnet.com/web2explorer/index.php?p=20&tag=nl.e539) and blog sphere (http://www.technorati.com/tags/web2.0) about the future of Web-based applications.
We here at Port80 give plenty of stock to this belief, but we also wish to throw a bucket of cold reality water on the hype fire. The stark truth of "Web 2.0"-based applications is that they are just as "dangerous" as they are powerful. By dangerous, we mean likely to fail. The beauty of the click-and-wait, batch-style Web 1.0 we use today is that failures are self-contained in those times between page loads. They are contained in moments the user is expecting them. These failures do happen -- network problems some place, lost packets, router hic-ups, server down, server busy, you name it.
With Web 2.0, failure can happen at many points which are not always well-defined to the user because there is much more communication going on -- and many more spots to get unlucky in performance. Couple this potential with the user's experience of desktop applications being responsive and reliable, and you have a recipe for dashed expectations with the first batch of Web 2.0 applications.
Now, this doesn't have to happen, there are plenty of ways to handle failures programmatically: exception catching with retries, talkback systems in code for active reporting of failures, diagnostic functions, and so on. You'll also probably want to add the redundancy to your server farm to make sure it can handle failures and not break any apps that are using your services. In short, you are going to have to start treating your servers and apps like Amazon does -- like a utility: it can't and shouldn't fail. Plan B has to be ready and waiting (and Plan C has to be thought about).
This utilization of the Internet isn't going to be easy nor is it going to be cheap. There are going to be many mistakes with people unaware of the app failures and those who over-engineer versus their load. And course there will be the peanut gallery out there filled with many people poking fun at those taking chances and failing. However, that seems the right way things should progress, as we saw this all before from the marketing angle during the .COM boom. It is tech and network folks’ turn now.
Game on,
Matt F.