I agree these are the biggest -- and serious -- concerns with the proposal. However, the status quo doesn't seem to be working too well; rather than discouraging frivolous or poorly-considered media types, it encourages people into the "x-" space.
I'm afraid this superficial analysis is fundamentally flawed, in that it ignores the fact that the rules have changed significantly over time. Back when MIME was first standardized the registration system was a total mess. The requirements were unclear, and even in the places where the require=ments were clear they were too hard to meet. Add to this the fact that IANA didn't have its act together at all, there was no web form, registrations submitted by email sometimes sat in a queue for up to a year before being acted on. All this has changed. The rules were substantially rewritten, the position of a MIME registration reviewer was created, IANA got new people in, reorganized the web site, and started dealing with requests in a far more expeditious fashion. These days it is simple to submit a request and responses often come back in a few days. But this cleanup took years to implement. And by then considerable damage had been done. The widespread use of of many if not most of the x- types you are complaining about was established almost a decade back under the old rules. In any case, now that the rules are what they are it is far from clear to me that there is a problem here that is worth solving. Are new x- types being invented today? The answer is probably yes, but my guess is that the rate has dropped considerably. And the people who are still doing it are likely ones who haven't caught on to the rule changes and are no more likely to switch to using a DNS tree than the existing vendor tree.
This is borne out when you examine mime.types files and Web browser configurations; deployed software and formats are ignoring the process quite freely.
And if you check the history you'll find that many if not most of these are very old.
I would strongly second this. Even very well-known types such as Macromedia Flash still universally go by names like application/x-shockwave-flash, which helps noone. There is no review of these names, and no means to prevent them clashing.
A situation which the instability inherent in domain names does little to help.
If using the DNS for providing unique names is seen as a bad idea, both because the names are volatile and because it is misusing a system that was never designed to provide GUIDs, then how about creating a separate registry of 'organisations' which can then manage their own trees? So, e.g., Macromedia could register */org.macromedia.* with the IETF, and then manage that namespace themselves. The idea of each MIME type carrying a parameter giving a URL with a human-readable description of the format is also well worth preserving.
We're going to go down that path we might as well use OIDs. Ned