Use cases: 1) A user building a reputation index 2) A service using this reputation data

Keybase.io supports: (example: https://keybase.io/mgwsoftware, https://keybase.io/cjb ) - nickname / username - "real name" - pgp pubkey (in ascii + fingerprint) - github [via gist confirmation] [ https://gist.github.com/cjb/10410941 ] - twitter [via tweet confirmation] [ https://twitter.com/cjbprime/status/648316716708167680 ] - bitcoin address - website(s) [via DNS record confirmation or file placement] - photo / avatar - news.ycombinator.com [ https://news.ycombinator.com/user?id=cjbprime ] - reddit [via post] [ https://www.reddit.com/r/KeybaseProofs/comments/3nei9i/my_keybase_proof_redditcjbprime_keybasecjb/ ] - devices?? https://keybase.io/cjb/devices ("sigchains" as a history chain: https://keybase.io/docs/sigchain )

Onename.io supports: (example: https://onename.com/gavin ) - email/username - location - text description - twitter - facebook - github - bitcoin address (ownership not verified) - website (ownership not verified) - pgp pubkey - photo/avtar from twitter

Bitrated supports: (example: https://www.bitrated.com/derek ) - email - text description - tags - SIN: https://en.bitcoin.it/wiki/Identity_protocol_v1 - others vouching for a user (!) - other reviewing the user - #bitcoin-otc - localbitcoins - coinbase - bitcointalk - twitter - reddit - facebook - hacker news - google+ - linkedin - github - stackexchange - angelist - paypal - fiverr - airbnb

Easily verifiable sources with Oraclize: - ethereum address - twitter (public tweet) - facebook (public post) [ https://www.facebook.com/yann.girard.1238/posts/1808311066062433?pnref=story ] - github (gist) - reddit (post on subreddit) - linkedin (you could add as project/organization "Iudex" + your linked address as part of the description) - bitrated/onebase/ - google? - youtube?

Decisions: - updating data will update in-place (e.g. we dont keep history in the contract) - changes history is based purely on blockchain history - history won't be shown in PoC - if we have enough time as a last milestone we'll support validated bitcoin addresses (checking via btcrelay and ecrecover) - secure linking/data processing and scoring are two separate conceptual parts; there can be multiple algorithms for scoring of the same source (e.g. twitter)

Process: 1) user creates a new iudex "account" via a transaction, gets a unique id assigned 2) posts that id on various mediums 3) initiates a "verification" for them via a transaction (and pays oraclize fees) 4) each verification triggers scoreUpdate 5) anyone can transact scoreUpdate (and pay the fees associated with it) 6) anyone can call data retrieval

Contract model 1 (bulky storage): 1) storage contract [SC] (has linked account idenfitiers) 2) verification contract [VC] (calls oraclize and stores the result in the SC) 3) scoring algorithm [SA] (uses data from SA, stores TYPE (e.g. twitter) and SCORE (the value) in itself) - there is a "special" algorithm with TYPE=ALL 4) scoring proxy [SP] (has the currently top voted scoring algorithms for each TYPE selected) 5) SA voting [SV] (-lets think about this-) 6) 7) lookup contract [LC] (stores the address of SC, VC, SP, SV?)

Contract model 2 (lean storage): 1) storage contract [SC] (has linked account identifiers, scores, timestamps for last score updates) 2) vertification contract [VC] (calls oraclize and stores the result in the SC) 3) scoring algorithm [SA] 4) .. voting, lookup, etc.

Reputation data should consists of: - identifier (first ethereum - "score"/"index" (yep I like this word) - linked accounts along with their respective scores a) twitter username (include verified tag) b) facebook userid/username c) linkedin username d) github username e) reddit f) onename/keybase/bitrated - milestone 2/3 - website? (place file in a given location, e.g. /iudex.key) - non trusted data: - email address - nickname/username - "real name" - "description" - "location" - pgp key - ethereum address(es) - date of last update

Actual storage data: a) Identifier b) Twitter/Facebook/Linkedin/Reddit/Github: username, ipfs hash, timestamp d) Array of ethereum addresses e) "hashmap" of non-trusted data (e.g. a mapping + another lookup, when we get there)

    Milestone 1: one "random text field"

f) Scores: twitter, facebook, linkedin, reddit, github, <overall>

Default algo for now - needs to be normalized somehow (range of 1 to 1.000.000): - Twitter: (followers-following)+0.25*likes - Facebook: friends, ? - Github: Contributions in the last year? followers? forks? - Reddit: link karma, comment karma - LinkedIn: connections, recommendations, skills votes

Any party should be able to call an "update" method to get fresh results back. Being it expensive the update is optional.

register() returns (bytes32)

- creates ID

sha3(msg.sender, block.number)

- stores `msg.sender` in `ethereum_addresses`

- returns ID

linkAddress(address)

- check if `address` belongs to any user

- store `address` in `ethereum_addresses`

unlinkAddress(address)

- check if `address` belongs to any user

- remove `address` from `ethereum_addresses` unless it is the only address

linkAccount(TWITTER, "@Username") unlinkAccount?

updateScores(ALL/TWITTER/..) updateScores(address, ALL/TWITTER/..) updateScores(uint, string, ALL/TWITTER/..)

getScores() returns (..) getScores(address) returns (..) getScores(uint, string) returns (..) //getInfo(TWITTER, "@ThomasBertani") = {'reputation_index': 99,}

getLinked() returns (..) getLinked..

getReputationIndex() return (..) getReputationIndex..