[dcc2] Question about the necessity of multi file sends

[codemastr]

> - - It doesn't require permanent connections
> Longer connection means higher probability of fault.
> A fault in a multi-file transfer makes it useless.

That is something I didn't think of, but it is a very valid point. If I have
100 sends being done through a multi-send, and the connection drops off
during send #2, 99 of the sends are not complete. Now lets assume that the
connection was lost just because someone accidentily kicked the LAN cable,
meaning, in 30 seconds the connection is back. With a multi-send, 99 failed.
With individual queued sends, only 1, maybe 2 are lost. Because it would
start sending #2, it would fail. It would wait for a timeout then terminate
the connection. It would then attempt to make connection for #3, and fail.
By this time, the connection should have been restored, and when the #3
timeout completes, #4 will start and should complete successfully. So in
that scenario, multi-send would have caused 99 files to be resent,
sequentially queued files require only 2 files to be resent. And really, in
such a scenario, it would be very easy to have file #2 and #3 automatically
resent.

Tom is however correct that the way DCC works now it would cause all 100
sends to be done at once. However, DCC2 != DCC. Why can't we standardize
some kind of queuing system? Just because the mechanisms are non-standard
now doesn't mean DCC2 can't standardize them!

It's a rough idea, but it could work.

-- codemastr