[Edited from comments made on the dotMobi blog]
As I understand it, .mobi is not as much about the fact a site is to be used on a mobile phone as it is to signify that the application or content is going to be suitable for usage in the context of mobility.
ilovetacos.com might be a corporate site with a rich, browsable ordering service.
ilovetacos.mobi, on the other hand, might be an interface designed to get you some tacos (with as few keypresses and dollars of data usage as possible), while you’re out and about.
I would have thought “finding out about tacos at leisure” is a totally different use case to “I’m on the move. Get me tacos now!”
It’s one web. Just two use cases. Right?
Additionally, if the browsers are able to do spectacular gymnastics to get full web sites running on a phone, then content developers may be lulled into a false sense of mobile comfort.
In fact, I would hazard thar ‘mobile applications’ are encapsulated by a philosophy far broader than fiddling about with markup and media-ruled stylesheets (on the server or on the client).
Could any browser turn a corporate taco web site into a location-based delivery service?
One final way of looking at this: if you could run a full version of FF2 or IE7 on your phone, might there not still be a good reason to visit the .mobi domain first (as you walk down the street)?