LibreOffice does footnotes and cross-references just fineSure, and it's a set of desktop applications that run locally on your own computer.
Did you realise you're supporting my position here?
The combination of cheap and easily accessible on any platform, without any lock-in whatsoever, is extremely powerful. Couple this with devices that are designed to be portable and always connected and you've got a killer that will cannibalize everything else.
We've seen thick client vs. thin client before, and no doubt we'll see it again. Today, however, we're still a long way from the utopia you describe.
Things like Google Docs / Google Calendar / Google Maps / Twitter / Wikipedia give me a hell of a lot more value than anything a desktop app could. Are you saying that you could build these in a weekend?
Well, you're obviously picking contrived examples, because most web applications aren't in that class. And of course no individual could literally build a replacement for those services in a weekend. However, it is interesting to consider some factors that would stop them:
1. Hardware resources and network bandwidth to serve so many users at once
2. Large volume of data behind the UI
3. Many import/export options, in some cases
Notice that of those, only the third has anything to do with how hard it is to write the software itself, and even then it's not core functionality. Providing compatibility with existing data formats is often extremely expensive in terms of developer effort, which is part of the reason that competing with the likes of Microsoft Office and Adobe Creative Suite is such a high barrier and why incumbents tend to change their file formats every five minutes.
On the other hand, if you were asking whether I could create a Twitter clone to share with friends at first and grow as friends-of-friends joined in, then sure, about a million people here really could write that in a weekend. The core functionality for most of the other services you mentioned could be implemented by a small team in times ranging from a few days to a few weeks, too.
The most interesting thing about this, IMHO, is that it means the core functionality for software with a completely different approach to solving the same problems might also be implemented fairly quickly, and isn't the factor limiting competition and the advancement of the technology. We're being held back by lots of other factors, but most of them are driven by commercial or marketing considerations, not the technology itself.