Comment 9 for bug 1157943

#6: SHOULD means that there are servers outwhere behaving like that. So we have to support both. Not that we can ignore should-clauses.
#7: Do I really have to comment that? I am not that good at explaining jokes, but lets try:
1. APT downloads a part of a file.
2. replaces the file on the server
3+4. Deal with the current file on the server which is from APTs view-point a new file as it has a part of the old file.
So in which way is it not dealing with changed content?

Both do not change the fact that you usually have a part smaller than the (new) file downloaded in which case the patch doesn't help (actually, I wonder how the patch is supposed to help at all as we still write the result unconditionally to the end of the file) – and it also doesn't change that I said I am not the one to talk to about https as I am not a big fan of "works correctly as the compiler isn't complaining"-patches and that is all I could do for https currently as I don't use https. (aka: I will try to help review working patches, but I am not producing one myself)