1 * min_owd TINT_NEVER is logged
3 v hints, data for non-existing ranges
4 v opens multiple channels to the same address
6 v RTT calculations need improvement (test)
7 v google-log is unnecessary
8 * reduce template use (peer queue)
13 v check hints agains ack_out?_
14 v check data against ack_in
15 v channel suspend/wake. 3 cong modes state machine - ???
16 * release hints for a dormant channel
17 * minimize the number of template instantiations
18 v Channel thinks how much it HINTs a second,
19 picker thinks which HINTs are snubbed
20 * files <1sec download : how HINTs are sent?
21 v dead Channels are not killed => cannot open a new one
22 (have a channel already)
23 v peers don't cooperate
24 * RecoverProgress fails sometime
25 v leecher can't see file is done already
26 v why leecher waits 1sec?
28 * file operations are not 64-bit ready
29 http://mail.python.org/pipermail/patches/2000-June/000848.html
30 * recovery: last packet
31 v no-HINT sending to a dead peer
33 v unHINTed repeated sending
34 v 1259859412.out#8,9 connection breaks, #8 rtt 1000, #9 hint -
35 mudachestvo, cwnd => send int 0.5sec
36 0_11_10_075_698 #9 sendctrl may send 0 < 0.000000 & 1732919509_-49_-45_-200_-111 (rtt 59661)
37 0_11_10_075_698 #9 +data (0,194)
38 0_11_10_575_703 #9 sendctrl loss detected
39 0_11_10_575_703 #9 Tdata (0,194)
40 0_11_10_575_703 #9 sendctrl may send 0 < 0.000000 & 1732919509_-49_-44_-700_-110 (rtt 59661)
41 v complete peer reconnects 1259967418.out.gz
42 * underhinting causes repetition causes interarr underest causes underhinting
43 * misterious initiating handshake bursts
44 v whether sending is limited by cwnd or app
45 * actually: whether packets are ACKed faster than sent
46 * uproot DATA NONE: complicates and deceives
47 v r735 goes to github; r741
48 * receiver is swapping => strange behavior
49 v on high losses cwnd goes to silly fractions => slows down recovery
50 v code the pingpong<->keepalive<->slowstart transition
51 v empty datagram hammering (see at linode)
53 * never back from keepalive syndrome (because of underhashing)
54 * HTTP daemon, combined select() loop
55 * range requests, priorities
58 v misterious mass packet losses (!data)
61 * check why the last HAVE msgs r not sent
62 * data is sent even if the client has stopped
63 * when initialized the piece picker might select hints from hint_out without knowing if the peer actually has it!
64 * IMPORTANT: trace(bin, range) reports bins out of range! Check if bug!!