Undo On Web

Ryan Stewart misses undos on the Web. He cites the example of Google Reader where he accidentally marked all feeds as read, and did not get an option to undo it.

There are many Web applications which support undos, and there are many desktop applications which do not. Which means it is clearly not a problem of the Web itself. They are more difficult on the Web as the underlying infrastructure and protocols are stateless, but we do have the technology today. I believe it is a subject of cost-benefit analysis. Does the undo provide enough value for the cost of effort and performance? Undo is not trivial, as it is not only about rolling back that action, it is about rolling back to an earlier state. Which means it has to undo the effects caused because of the action, which can be quite expensive. Sometimes other options might work better, e.g., using a trash can to retrieve deleted files or emails.

And then there are some actions which cannot carry an undo with them. For example, sending emails with empty subject or sending an IM message or an ecommerce transaction, which I call the points of no return. At the most you can take another action that negates the preceding action (like canceling an order), but it is still not an undo.

The problem here is because it is difficult for the computer to find out whether you are taking the action intentionally or accidentally. Of course undos are the key to making software more tolerable to your mistakes, but they do not come cheap. If you see undo missing for an action from an application, it is quite possible that it has failed the cost-benefit analysis.