There is no need to be defensive, nobody is attacking anybody. This is just another "best practices" thing.
update: just to add one more note: end-users don't care if the person who built the product like/dislike implementing an undo feature, end-users care if it's easy for them or not.
There's a somewhat worrying trend with applications such as Facebook and Google Apps where software is more concerned with making users believe it's doing what they need it to do, and less concerned with doing what users need. I'm not saying this necessarily crosses the line - if I have the space on my server I'm storing lots of incremental backups. But there's always a trade-off. And some people might rather be assured that when they hit delete that it goes away forever than that they never lose anything they meant to keep.
Outside of anything real time, most things have the capability of having an undo element added to them to some degree. Even then, real time actions can still be undone (credit card transactions, for example).
I agree with the author that habitual responses to dialogue buttons can be dangerous, so I suggest considering this on a case-by-case basis - not just as to whether it's worth the money, but whether it will significantly reduce user distress.
Yes, it's difficult to bootstrap undo functionality. But look at mobile apps, they never ask "are you sure you want to leave?", they just make sure that whenever you come back - everything is exactly like it was before. Users EXPECT this behavior and developers feel obligated to implement it.
Concerning deleting by accident - I think it's just that we have bad software usage/development norms. Who knows, maybe this will improve with time.
Undo functionality is like that. You have to start with some sort of ability to make it easy to create your "actions" in some way where every one of them can be reversed, and in some cases the OS itself makes that very hard. How do you undo "close program"? Like you said, but didn't quite spell out, with a fundamentally different OS where suddenly it's easy. You can make a Microsoft Windows program that will automatically open in the same state you left it in, but first of all, it's much harder than it is on a phone, and second, I very much hope that you thought of that from day one because otherwise you're probably in for some pain just refactoring your program to the point where it is possible....
Ideally you would work in some language that makes undo that easy, but it's really hard because so much of the system is fighting you. I've toyed with this but UIs themselves are fundamentally not undoable so you burn a lot of time just telling the UI how to roll back. It really needs to be fundamentally woven into the entire system and that will probably be decades, if ever.
I know your context was about something a little different but just to be nitpicky, most iOS apps we develop prompt the user if they want to leave the app and launch Google Maps to get directions and/or thank them for wanting to donate - Safari will be launching shortly and/or tell them to wait while we launch the e-commerce page.
You can cancel the launch. But you can't undo it. It is thermodynamically impossible.
When I clicked on "Empty Trash now" all I got was this message: "All messages in the Trash have been deleted forever."
I’m not seeing any assumptions about the difficulty of implementing an undo feature in that sentence (except that it’s not impossible and that some companies have been capable of implementing it). Where do you find those?
I think your prejudices got the better of you.
Are you saying that Illustrator would be just as good without undo?
Illustrator is clearly better with undo, but I don't think basecamp, for example, would be.
Other things, you want to delete only occasionally, and may be of high importance, like a character in a video game. this is when you want an alert. If it's not something you do all the time, it's not habit forming ( in the context of this particular game). It adds no complexity to the ui and has nearly no implementation cost, yet it helps protect your important data.
Now you add an undo button and you've gone from two actions to three actions. Furthermore, does the undo apply just to when you delete characters, or can you undo the progress you just made with a character? If it's the former, then your undo is inconsistent. If it's the latter, if you accidentally undo, then you want a 'redo' so you can't lose your new changes either. So now you've doubled the number of options the user can take. And how often are you going to delete a character anyway?
Or, you can take the approach games do, which is simply ask you 'are you sure you want to delete this character?' Given this response is distinct from the 'continue game' action, which would display a loading screen, players are likely to notice the prompt. Why complicate things?
As for consistency in a game, menus and the games UI rarely share the same structure. It's also generally except that you don't put up a confirmation box for the player to level up.
Honestly, how many actions in a game does it make sense to undo, unless you're playing Prince of Persia: Sands of Time? I'm a big proponent of 'make everything undoable' but you are stretching it way beyond its use here.
For instance, even a simple message exchange, which has no data interdependency issues, an undo feature requires some kind of design trade-off. Either the application lies to the user and delays the sending of the message to allow for an undo period, or it allows the message to be delivered to the other user, but then allows the delivery to be un-done. However, in the later scenario, if the receiver reads the message before it is "unsent," then you aren't really providing a reliable undo facility.
A typical designer point of view on the world. Has the author any idea of how much time (therefore money) it can take to implement Undo features across the board? I don't think so. Successfully implementing Undo functionality can be a daunting task. The question simply comes down to; is it worth the money investing in this feature? For most applications it simply isn't, no matter what mister visual UX dude thinks or says.