plan: reputation group

[0xtsukino]

TODO:

• sunset old interep groups (cannot join this group anymore, but should work for anyone already in the group)
• decide threshold for twitter/reddit/github membership tiers
• implement reputation verification logic (gatekeeping for joining the group)

zk-group permissioned group for twitter/reddit/github

• at some point ew can use zk-group to do rep verification too
• for now we should just keep this group centralized and private on zkitterd (indexer)

dao, and github contributors

• for now we should just keep this group centralized and private on zkitterd (indexer)

• on-chain: eth2 validators (plan later)

[0xtsukino]

When inserting new member, let's keep track of usernames

[sripwoud]

For the "decide threshold for twitter/reddit/github membership tiers" step:

0. Decide which stats to collect
   Selection criteria:

   • Should be publicly available: we need to be able to scrap this data! So we exclude private data.
   • API to fetch data points should exist (I don't want to scrap HTML😅)
   • should "contain" useful information(i.e. can be used in a reputation "calculation"): numerical info is better and urls like e.g profile image are useless

   Suggested stats to include in reputation score:

   twitter	GitHub	Reddit
   followers count	total stars earned	comment karma
   following count	total PRs	link karma
   creation date (account age)	total commits	creation date (account age)
   botometer score	total issues opened	premium subscription?
   verified	followers?	coins?
   	sponsorships?	linked identities?
   	total contributions?
   	total received stars?

   Difference with current stas included in current interep thresholds:

   • github: exclude "proPlan" as it is not available in public data (only on authenticated profile data).
   • twitter: keep same, adjust thresholds
   • reddit: to be defined (are coins, linked identities and premium subscription stats publicly available?)

1. Collect public data samples for each oauth provider: min ~1000 data points for each.

2. Bin data according to a given distribution to find thresholds
   Find the bin edges that will match a given distribution: e.g bronze/silver/gold - 60/30/10. These bin edges will be our thresholds.

[sripwoud]

@0xtsukino @AtHeartEngineer what do you think?

@cedoor For the reddit stats: premium subscription, coins and linked identities are used by interrep. Do you know how to fetch this data? Is it even available by fetching public user data with the reddit api?
I am a bit struggling to use the reddit api (it is not very well documented) to collect data, so I might just end using a public dataset like this one.

[cedoor]

Hi @r1oga!

I am a bit struggling to use the reddit api (it is not very well documented) to collect data, so I might just end using a public dataset like this one.

Isn't that dataset old? Or is it updated?

Do you know how to fetch this data? Is it even available by fetching public user data with the reddit api?

Interep is using this API: https://github.com/interep-project/reputation-service/blob/main/src/services/reddit/index.ts I don't think you can fetch user data without their OAuth2 token

[sripwoud]

Thanks for the info.
yes that dataset is probably a bit old.

cant fetch user data without Oauth token

Oh so it would mean you can only fetch the data of an authenticated user?!

I am going to try this one https://www.reddit.com/dev/api/#GET_api_user_data_by_account_ids
to fetch user data from randomly generated ids.

[cedoor]

Oh so it would mean you can only fetch the data of an authenticated user?!

It's something we should test, but I think you need a token to fetch some data, maybe not everything though.

I am going to try this one https://www.reddit.com/dev/api/#GET_api_user_data_by_account_ids to fetch user data from randomly generated ids.

Which data does it return?

[0xtsukino]

I am going to try this one https://www.reddit.com/dev/api/#GET_api_user_data_by_account_ids to fetch user data from randomly generated ids.

how about:

• sample random posts from this api: https://www.reddit.com/dev/api#GET_random, get creator
• query user data from this api: https://www.reddit.com/dev/api#GET_user_{username}_about

seems like both requires an oauth token, but can be used to query about any user/listing

[sripwoud]

Thanks, will try that too.

[sripwoud]

@0xtsukino
For the gh data here are intermediary results and first new thresholds suggestions. (In the interep org as I had started this work there).
interep-project/interep-groups-eda#7

Instead of the 60/90 quantiles like I had suggested above, 90/99/99.9 seem to be better, at least for GH data.

[sripwoud]

update (for sprint planning today):

• sunset old interep groups (cannot join this group anymore, but should work for anyone already in the group): not done, I'll discuss with cedoor how to do that

• decide threshold for twitter/reddit/github membership tiers: 80% done. Reddit and twitter done, github data collected but need to clean/crunch it and run some simulations.

• implement reputation verification logic (gatekeeping for joining the group): not done (part of the services are ready, like this API which return the data for the GH groups)

• zk-group permissioned group for twitter/reddit/github: still need to sync more in details with andy and cedoor

• dao, and github contributors: not complete, partly done, see https://github.com/zkitter/groups

• eth2 validators (plan later): we can reuse some endpoint I build for https://github.com/privacy-scaling-explorations/e2e-zk-ecdsa. see here the /beacon endpoint

@0xtsukino