the thing actually works pretty good now
This commit is contained in:
parent
0f431621b9
commit
5c5cf3c24a
2 changed files with 26 additions and 5 deletions
24
TODO.md
24
TODO.md
|
|
@ -8,5 +8,27 @@
|
|||
- [ ] per-file stats
|
||||
- [ ] per-peer stats
|
||||
|
||||
- [ ] slow peers cause slowness in the end, need the "end of game" algorithm
|
||||
- [ ] will require implementing cancel message
|
||||
|
||||
someday:
|
||||
- [ ] cancellation
|
||||
- [ ] cancellation from the client-side for the lib (i.e. stop the torrent manager)
|
||||
|
||||
|
||||
# concurrency
|
||||
it's fucked up now, so need to rethink.
|
||||
|
||||
Sequencing:
|
||||
- when the peer sends bitfield
|
||||
- update its bitfield
|
||||
- this can only happen at the start. But it can also NOT happen at all (if the peer wants to download)
|
||||
- so we actually cannot use it as a trigger to start the "uploader" of it (where we upload to it)
|
||||
- however both this and "have" we can use as a trigger to start the "downloader" part of it
|
||||
- when the peer sends "interested"
|
||||
- when the peer sends "have"
|
||||
- update its bitfield
|
||||
|
||||
"peer downloader":
|
||||
- if started, means there was some initial interest
|
||||
- if we are unchoked:
|
||||
- fetch new pieces forever from the queue, send requests
|
||||
Loading…
Add table
Add a link
Reference in a new issue