Your idea (finish chunk before shutdown) makes no sense. Please please please take your time to understand following concept.
Case A - commonly available file.
[...]
Conclusion - no impact for any given client.
no objecton - if a file is widely available, there's no real need concentrate on details...
Case B - rare file (#2 doesn't hold)
Property #1 (file is rare) - there are many clients in queue
For rare files situation is different since dropped client will increase time in queue for .
i'm confused, could you explain a little more the queue point?
But, due to #1 it doesn't matter whether one of many waiting clients got the chunk - on average everyone is screwed.
well if one or two of the clients got the full chunk they can continue spreading it until I'm back online
But, most important, both case A and B lay on the assumption #3 T>>t, there is a not irrelevant percentage of people that use file sharing programs (T) only one or two hours a day, not more tha four hours anyway, considering an average "t" about 30-40 minutes (that is absolutely sensible in Italy and, I think, in Europe), you can easily see that the impact is not irrelevant in these cases