I have great respect for what you guys have accomplished, and I am very grateful, but I don't really understand this idea that resumes will naturally get better as the infrastructure is improved. Sure, if Limewire supported hashing, etc., then resuming downloads could be more robust -- assuming they worked at all. But it seems to me that they DON'T work at all, ever, and the real problem is not architectural but just that there are *specific bugs* in the resume implementation that prevent it from working. (e.g., when Limewire resumes a download, even automatically, it usually restarts it at the beginning in another file with a different name, or restarts at the beginning of an already partially downloaded file, overwriting what was already saved.)

So you can improve the architecture as much as you want but it seems to me that this won't be of any use unless the bugs are fixed. Example: requerying was introduced in 2.2 and unfortunately this feature is essentially worthless because when a partially downloaded file is automatically requeried and restarted, it doesn't resume properly. I don't see what the point is of adding features that can't be of any use until more fundamental bugs are fixed.