How would you implement such a system? There are lots of factors, including trademarks (many overlap, e.g., Linux the OS and Linux the detergent), corporate interests and personal interests. There is also the problem of how useful some of these TLDs will actually be, especially given how people are used to the existing system.
In the UK, the .plc.uk, .ltd.uk and .net.uk domains have to be vetted before being registered - this means that it is very hard to cybersquat in these domains, but the .co.uk and .org.uk names are still up for grabs by anyone. Should a similar system be used for these new TLDs (e.g., you have to prove that you are a retailer before you can own a domain with the .shop suffix), or should we give even more money to lawyers to sort out the problem?
There are also many other problems - for the personal domains (.psn or .fam) how do you arbitrate between claims? Is the first-come-first-served model appropriate for these types of domain? Why should one person get smith.fam, when there are several million people out there who have an equal right?
The article was interesting, although it didn't go deeply enough into the issues for my taste. (Granted, this was a news item and not an opinion piece like this one.) This is certainly an issue that will affect all of us. However, the article failed to ask what I consider to be a fairly obvious question: What's to keep people from registering domains in other countries? Foreign companies register .com addresses all the time -- and Tonga seems to be a (semi-)popular country domain. (Example: http://surf.to/mysite/, http://click.to/somewhere/)
I've also seen ads advertising, "Register your .cc domain now!" But no matter what they do, .com will be (at least for the forseeable future) the holy grail for domain registrants, mostly due to a general public ignorance of the fact that there are TLDs other than dot-com. So nothing's going to really "change" if they only introduce new ones.
What's really needed, IMO, is a couple of ideas that they started to postulate in the article, but never really talked about at length. The way it's set up, the canonical meanings for the TLDs are .com for commercial organizations, .net for network operators/providers, and .org for "anything else", usually assumed to mean non-profit organizations. (.edu, .mil, and .gov don't really count as they're not available to the general public.) In practice, as you well know, it's a free-for-all; people register all three for just about any purpose they wish. http://www.harrybrowne2000.com/, for example, isn't really a traditional commercial organization (although one could argue that that's the most appropriate domain for a political campaign), and would be better served, perhaps, with a .org domain. (In all fairness, they do have the same .org domain which points to the same server; actually, they have no less than four different domains. Drop the '2000' for the other two.) Slashdot.org, on the other hand, is not an NPO (although it was the most appropriate TLD at the time); they now sell banner ads on their site and make money doing it (theoretically ;) ) and have been acquired twice. This type of dilemma would apply also to k5 when it gets big.
I personally like the idea that, for example, you'd be required to prove you're a merchant before you'd be able to acquire a .shop domain, for example. I'd also like to see a domain called, for example, .npo, for which you'd be required to prove that you're a non-profit organization. (In that case, the current .org would work rather well as it is, since it'd kinda resemble a .misc domain.)
As for a .per or .indv (for personal/individual sites), the site brought up the fact that these would be rather difficult to arbitrate, since several hundred thousand people, for instance, would all have equally valid claims to the smith.per or jones.per domains. The article brought up this problem, but didn't suggest a solution. (Which, again, is perfectly valid since it was a news article and not an op-ed piece.) One possible solution might be to register serveral hundred thousand of the most common names in a surname.per format, and let people register givenname.surname.per. This could still cause problems, though; the most famous Robert Smith, for example, is the lead singer of The Cure, but there are thousands of other Robert Smiths in the United States alone.
Only time will tell how these issues will sort out, but I suspect it will be a combination of convention, and unfortunately, legislation and litigation.