4 * imposed HINTs are terribly broken, resent for the data in flight
5 * check ACK/HAVE redundancy
6 * HAVE overuses find_filtered
7 * set priorities on ranges
8 * small-progress update problem (aka peer nap)
9 guarantee size of updates < x% of data, on both ends
10 * pex is affected by peer nap
11 * how will tracker aggregate pexes?
13 * HAVE ALL / HAVE NONE
14 * aggregate ACKS (schedule for +x ms)
15 * channel close msg (hs 0) # Arno: indeed, there appears to be no Channel garbage collection
16 * connection rotation / pex / pex_del
17 * misterious bug: Rdata (NONE)
18 * ?amend MAX_REORDER depending on rtt_dev
19 * Tdata repetitions bug
22 * move to the.zett's binmaps
23 * optimize redundant HASH messages
24 * move to rolling HAVE queue
27 * initiate RTT with prev RTT to host:port
34 * file/hash-file re-open in read-only mode
35 * no cache recheck, failure-resistant
37 * unified events/callbacks
39 * Transfer(fd) constructor
40 * think of sliding window(s)
41 * the ability to sniff file without downloading
44 * all-swarm performance stats
45 * run chained setups (cmd line protocol subsetting)
46 * implement: multiple swift instances per server
47 * run thousand-daemon caching tests (use httpgw)
48 * use a dedicated tracker
49 * add NATs to the setup
50 * recover mfold.libswift.org
57 * NAT type detection => need peer identifiers (x100 amplification)
60 * integrate multi-peer changes by Jori
61 * do global swarm stats
65 * Sha1Hash constructor ambiguity
67 * think of using HTTP (?) as a fallback
68 * add header/footer, better abstract to the draft
69 * Gertjan: separate peer from channel? cng ctrl per peer ?
70 * packing hashes into a single datagram (tracking 1000s)
71 * partial channels / lightweight channels
74 * 6 degrees of sep = 3-hop TorrentSmell
75 * 60% immediately not connectable