The main services a publisher provides are final selection and then editing. The selection is done in the first instance by literary agents; most publishers now will not accept a submission except from an agent.
This selection has good and bad points. It weeds out material which is unpublishable, but it also militates against work which is arbitrarily deemed uncommercial. On the one hand, the reader is spared a lot of slush. On the other, the reader misses out on all kinds of new, original and possibly wonderful books. For example, most of Kafka's output remained unpublished at his death, and we all know stories of eminent modern authors who suffered years of rejection.
Editing is vital. Writers vary greatly in their need for it. Some need major restructuring of their work or have at best an imperfect grasp of grammar and usage, while others need only a light touch. Editing is a skilled and expensive process and I would suggest that sometimes the editor can be almost as important as the author.
Publishers are also responsible for promotion, but increasingly this is targeted at the books of authors who are already successful, with little or nothing left in the budget for newbies or mid-listers. My experience, when I was published in paper, was that the size of the promotion budget was directly related to the size of the advance. The very best form of promotion is actually free: it is word-of-mouth.
Thus, if one is to cut the publishers out of the loop, one must (a) select and (b) edit.
Both of these would be taken care of by a central site where self-published ebooks were reviewed. The site could be funded by advertising and by charging authors a small fee for each listing, or each click-through, or something of the sort. Professional reviewers should be engaged; their reviews should be rated and commented on by readers. In addition, readers should submit reviews, and these too should be rated and commented on, which would effectively stop authors spamming their own work.
The database of titles should be searchable by all sorts of criteria, such as genre, rating, price, and the Amazon-style "if you liked that then try these". See also
a piece I posted at TeleRead in October 2008.
Good books would rise to the top by a process of natural selection. I say "good", which is not necessarily the same as "commercial". If a book were badly edited it would score badly in the reviews, putting selection pressure on the author to get his act together in that respect.
Does anyone else think this might work?