|
|
|
- Take care of posts using older public key when key is replaced.
|
|
|
|
|
|
|
|
notes: not very difficult, GetTransaction must receive a maximum block number to search the
|
|
|
|
transaction (we get this from post["height"]). another txIndex should be set to speedup lookup
|
|
|
|
(key in db includes the number of the block that changed tx so previous one can be found).
|
|
|
|
pseudocode:
|
|
|
|
getTxIndex( key = "userX" ) => block h contains this tx;
|
|
|
|
while( h > max_h )
|
|
|
|
getTxIndex( "userX_h" ) => block h contains the previous tx
|
|
|
|
|
|
|
|
- Count UTF8 chars in acceptSignedPost to proper limit the 140 characters.
|
|
|
|
|
|
|
|
- Encrypt user_data (which contains all DMs)
|
|
|
|
|
|
|
|
- Test wallet encrypt to see if it still works from original bitcoin implementation and what
|
|
|
|
are the implications to our code.
|
|
|
|
|
|
|
|
- Rescan directmessages after importing a privatekey (importprivkey)
|
|
|
|
|
|
|
|
- Check libtorrent's limitation on the number of pieces (max_pieces in piece_picker.hpp = 1<<19)
|
|
|
|
Since post number is constrained by max of 288 posts per day in average, that means we have 5 years
|
|
|
|
to think about it (for the really heavy users).
|
|
|
|
|
|
|
|
- Besides increasing the maximum number of pieces, a more pressing issue to save bandwidth and
|
|
|
|
torrent download time would be to define the first piece to download/store locally. People don't
|
|
|
|
need to maintain the entire post history for everybody they follow, they could just keep the last
|
|
|
|
ones. This has to be implemented.
|
|
|
|
|
|
|
|
- Move all crypto to javascript, store only encrypted version of the privatekey (which would be
|
|
|
|
decrypted only in browser memory). getposts may obtain all DMs encrypted to browser, another
|
|
|
|
newpostmsg needs to be provided to receive posts with signature field added.
|
|
|
|
|
|
|
|
- Store a dht resource "publickey" containing not only the public key itself but also information
|
|
|
|
needed to validate it by a lightweight client. That includes: block hash, block height and partial
|
|
|
|
merkle tree inside that block. This resource propagation cannot be sent right after user
|
|
|
|
registration for obvious reasons (no block yet, other nodes wouldn't accept the signed dht put).
|
|
|
|
|
|
|
|
- Discuss and implement the acceptable level of spam per day (priorizing localization).
|
|
|
|
|
|
|
|
- Implement the mention forwarding mechanism discussed in the paper so user don't need to do polling
|
|
|
|
and can also be sure to receive all mentions.
|
|
|
|
|
|
|
|
- Implement hashtag "storage-less" torrents for post distribution.
|
|
|
|
|
|
|
|
- Define expiration policies to dht stored values. Currently all keys are refreshed every hour which,
|
|
|
|
according to previous bittorrent research, would be enough to keep data available forever (with high
|
|
|
|
probability). twister also persists keys to disk. As userbase increases, old post storage and
|
|
|
|
unreliable multivalued keys should better expire. Since those posts include the height and time, a
|
|
|
|
policy may me defined.
|
|
|
|
|
|
|
|
-
|
|
|
|
|