Well, actually, I don't hate them at; i just hate not having any tools to tell me what exceptions could get thrown on the calls I make.  Of course I realize how that might not be possible to compute given that you really don't know what could happen with a virtual call, however i still think a lot of useful information could be given.   You could certainly start by marking any method that actually throws an exception (which escapes the method) with the information about that exception.  You could then take the transitive closure to determine the extent that those exceptions could reach.  You could use that information to make sure that the doc comments for methods didn't forget to mention those exception, and you could also use preexisting <exception> tags to augment the scanner for methods that could not be determined automatically.

I'm also not convinced of the problems that Anders' raises here.  However, I don't have enough experience with large scale applications to know whether or not exceptions work well in those situations.  I do know I've seen checked exceptions horribly abused (i.e. catches that, innapropriately, swallow exceptions.  every methods decorated with "throws exception").  Of course, in the past I've stated that sometimes it's good to not have a features if its going to be abused and people have lambasted me for that.  Now it appears I'm going to be hypocritical by saying that I think in this case we should have a feature even if it can be abused.  However, i think that because it seems like it's so beneficial, and even if it's abused it's no worse than our current system.  Of course, I do understand Anders when he says "Once a better solution is known—and trust me we continue to think about it—we can go back and actually put something in place. I'm a strong believer that if you don't have anything right to say, or anything that moves the art forward, then you'd better just be completely silent and neutral, as opposed to trying to lay out a framework."  However, it's been a few years since the inception of C# and no perfect solution has been found, so maybe it's time for just something that makes things better.

I'd really like to work on something that could go and analyze existing code/metadata and garner this information for the user.  I think it would be helpful and it would certainly be great for learning how to read and examine IL instead of just producing it.  Of course... now I just have to find the time to do so :-/