The TLD issue is one I'm still trying to resolve in my head. I guess if I have a point to contribute, it's that there is no reason for a new TLD unless you are unhappy with the allocation policy that exists in existing ones. ".museum", for example, has a set of requirements you have to meet to register in it. But I'm not convinced that this sort of extra hurdle actually adds value to the name.
If we have enough such "set-asides" to keep everybody happy (although whether that's possible is another issue), then there are probably going to be enough so that their mnemonic value is too small to be useful. If I'm looking for the San Francisco Museum of Modern Art, do I look at sfmoma.com, sfmoma.org, sfmoma.museum, moma.sanfrancisco.museum, sfmoma.art, sfmoma.pictures, sfmoma.sculpture, sfmoma.photo, sfmoma.mus, moma.sf.ca.us, etc.? (Although I do like what .museum does--- if you guess wrong, it dumps you into a list you can search through.)
So, if we can't make people who want to try random URLs (and those that want to catch their attention) happy, does "sfmoma.museum" give more confidence in the result of a search engine than "sfmoma-museum.org" does? Perhaps.
Anyway, I hope to have an argument which make sense in the paper I'm writing.