build content blocking system

[hburgund]

Per Apple's "new" requirements, we need to be able to let users specify content they want blocked i.e. never added to their stream. There are two types of content:

• assets by asset_id
• assets by created_by user [ asset.id -> asset.session_id -> session.device_id -> userprofile.user_id -> auth_user.id ]

In other words, users can block a specific piece of content or they can block all content by a certain user.

Proposed Solution

We want to keep the ability for anonymous use, which should be possible. I envision creating a models.BlockedContent model that stores information on what content has been blocked and then playlist filtering must take this into account when it generates playlists. Perhaps we can use the banned assets feature as an example, though that is on a session-by-session basis, not an overall ban.

models.BlockedContent could have the following fields:

• id
• blocker - the user doing the blocking
• blocked_asset_id - ForeignKey to Asset model
• blocked_username - ForeignKey to User model (not sure if rw_userprofile or auth_user is better here)
• session_id - just for good measure to track things

It seems better to me to store all blocks in one model rather than storing asset blocks in one and user blocks in another. Perhaps if a record contains a blocked_username with no blocked_asset_id then the assumption is that all content produced by that user is blocked.

With this data in place, every time a new session is created, the session user is known, so this model can be polled for all blocked content and then that blocked content can be stored in a temporary "session blocked list" of some sort that is cross-referenced each time the list of available assets is created or updated. Blocked users would be converted into a series of blocked assets to create the blocked list so that the list is simply asset_ids.

API UPDATES

• POST /api/2/blockedcontent/ - params [ user asset_id block_type session_id] (block_type would be "asset" or "user" to indicate a single asset block or all assets by user of submitted asset_id)
• GET /api/2/blockedcontent/?user=<username> - retrieves all blocked content for specific user

Naming

I'm not sure "block" is the best terminology here, but it may be. "ban" could be better to avoid all the other meanings, technical and not, of "block". We already use "ban" for time/proximity banned assets; not sure if consistency there would be good or confusing.

[hburgund]

Different Approach using Vote Model

Another possible solution has come to mind. Rather than creating a new BlockedContent model, perhaps we could make use of the already existing Vote model as that was built with extensibility in mind. The purpose of votes is for users to register their "opinion" about a specific asset, so if that "opinion" was block, maybe that solves our issue?

The Vote model has the following fields currently:

• type - currently like flag or rate
• asset_id - ForeignKey to Asset model
• session_id - ForeignKey to Session model
• value

asset_id and session_id overlap with the proposed BlockedContent model and the latter can yield blocker, though I would argue that it is probably easier to just add a voter field rather than having to traverse models all the time. If we instituted two new vote.type values - block_asset and block_user, that might be enough to handle our needs.

I like using what we already have rather than adding an entirely new model, and I do think that this use case is aligned with the definition of "voting". BUT I don't want to shoehorn functionality into a place where it shouldn't be just to save adding a new model if ultimately it would make things more complicated.

@eosrei @jslootbeek any opinions?

Changes TODO

• add new vote.type options - block_asset block_user
• add vote.voter field - ForeignKey to auth_user model (settings.AUTH_USER_MODEL)
• update GET and POST api responses
• ensure all vote calls put voter into new field
• build blocking system to filter playlist by blocked assets and blocked users per session's user