[Implemented] Fully Disable Discover & search results from Streaming Services (More Ways To Watch)

ALL:

I have answers for you.

IF I understand it all correctly:

  1. The decision was made to provide the feature

  2. As it was sketched out, a tentative schedule was created

  3. Somewhere that tentative date became a fixed deployment date

  4. Development was harder than expected (which is very common)

  5. With obligatory date at hand, the feature was enabled. (you all already had the code but it was hidden behind a feature flag)

  6. The testing, which had been occurring throughout, wasn’t complete prior to launch date.

  7. The diversity of every found all the glaring problems almost instantly.

  8. The team looked at the feedback/reports was coming back and jumped into action (which is before I got involved)

  9. Now that I’m involved, I’m serving to take forum interaction load so they can work uninterrupted.

Problem breakdown:

  • Metadata contamination and handling thereof
  1. There exists, as you all have seen with your own media, metadata which has no tagging / rating on it. The actual values, as has now been learned can be blank or null.

  2. There also exists “MA” (mature audience) tagged metadata which is incorrectly tagged and should be much higher.

How did this data get into the otherwise pure databases ?

Somewhere within the Plex userbase, someone has this “inappropriate for children / NSFW” media and added it to their server. (It’s their server – their choice)

When they added it to their Plex server

  1. The title got sent upstream to obtain the match
  2. Upstream didn’t have the data needed so it went out to search.
  3. It found the data and ingested it.
  4. The data was then sent to the user to populate the user’s library section (normal operation).

This is where things went sideways.

  1. The data didn’t have all the necessary tags.
    (There’s no way to discern in an automaton if it’s the right tag value).
  2. It was allowed to go into the database

Now for the collision of Discover and this metadata.

  1. Discover expects tags / ratings

  2. Searching expects tags / ratings

  3. The degenerate case of tag matching (when a tag is NULL/empty) occurred when the search keyword encountered data with a null tag. (instant match).

  4. You type two words ( I won’t type them here) to search and get back XXX material – BECAUSE it had no rating / tag.

Where are the errors here:

  1. The user-facing side didn’t check for the completely degenerate case. It shouldn’t have to worry if the upstream database is always pure. It WILL have to worry about that if you have your own metadata for non-published media (not in the internet databases).

  2. The Plex databases had ingested the untagged metadata from performing searches for users.

What’s happening:

Metadata

  1. All the metadata is being cleaned out

  2. The metadata retrieval & ingestion will now check before ingesting.

User Interface

  1. The controls everyone wants was already in progress.

  2. Based on feedback, the design is being enhanced.

What can we expect?

  1. Discover will stay. It will not be disabled/pulled back (business/contractual obligations)

  2. We will get the control. I do not know what form it will be (there are no mock-up graphics yet)

  3. The metadata database and how Plex’s side works is getting smarter & “tighter” .
    (a lot of the worst offenders have already been removed)

  4. The metadata search functions (which is what you see) have already been ‘tightened’ about what results they return for searches.

  5. I don’t have a date for GUI change deployment. I will share as soon as I can / have test changes in my hand.

This is what I have after 36 hours involvement.

EDIT: OPT-IN / OPT-OUT will be part of the GUI changes which I will share when I see / have something in hand.

22 Likes