oEmbed providers.json parsing

[jonas-w]

Description

As I wanted to check if #17462 can be fixed by using oEmbed I stumbled upon parsing issues in the oEmbed config parser. I used the official provider official providers.json from https://oembed.com/providers.json

1. The config parser throws an error and hinders Synapse from starting, when a provider in the providers.json does not contain the "schemes" key, IMO this should just throw a warning and not error out.

   When a provider in the providers.json does not have a 'schemes' entry it can't be used directly in Matrix, as there is no URL pattern to match, these entries without 'schemes' are seemingly only there to signal that these websites have a discoverable oembed provider.

   You can see which providers don't have schemes via this jq command: jq '.[] | select(.endpoints[].schemes == null)' providers.json

   This is the current output of this command:

   {
     "provider_name": "Beautiful.AI",
     "provider_url": "https://www.beautiful.ai/",
     "endpoints": [
       {
         "url": "https://www.beautiful.ai/api/oembed",
         "discovery": true
       }
     ]
   }
   {
     "provider_name": "Box Office Buz",
     "provider_url": "http://boxofficebuz.com",
     "endpoints": [
       {
         "url": "http://boxofficebuz.com/oembed",
         "discovery": true
       }
     ]
   }
   {
     "provider_name": "eduMedia",
     "provider_url": "https://www.edumedia-sciences.com/",
     "endpoints": [
       {
         "url": "https://www.edumedia-sciences.com/oembed.json",
         "discovery": true
       },
       {
         "url": "https://www.edumedia-sciences.com/oembed.xml",
         "discovery": true
       }
     ]
   }
   {
     "provider_name": "eduMedia",
     "provider_url": "https://www.edumedia-sciences.com/",
     "endpoints": [
       {
         "url": "https://www.edumedia-sciences.com/oembed.json",
         "discovery": true
       },
       {
         "url": "https://www.edumedia-sciences.com/oembed.xml",
         "discovery": true
       }
     ]
   }
   {
     "provider_name": "GloriaTV",
     "provider_url": "https://gloria.tv/",
     "endpoints": [
       {
         "url": "https://gloria.tv/oembed/",
         "discovery": true
       }
     ]
   }
   {
     "provider_name": "Matterport",
     "provider_url": "https://matterport.com/",
     "endpoints": [
       {
         "url": "https://my.matterport.com/api/v1/models/oembed/",
         "discovery": true,
         "formats": [
           "json"
         ]
       }
     ]
   }
   {
     "provider_name": "PingVP",
     "provider_url": "https://www.pingvp.com/",
     "endpoints": [
       {
         "url": "https://beta.pingvp.com.kpnis.nl/p/oembed.php",
         "discovery": true
       }
     ]
   }
   {
     "provider_name": "Rumble",
     "provider_url": "https://rumble.com/",
     "endpoints": [
       {
         "url": "https://rumble.com/api/Media/oembed.{format}",
         "discovery": true
       }
     ]
   }
   {
     "provider_name": "Typlog",
     "provider_url": "https://typlog.com",
     "endpoints": [
       {
         "url": "https://typlog.com/oembed",
         "discovery": true
       }
     ]
   }

2. The Parser also errors out for this spotify scheme: spotify:*. Again IMO this should probably just throw a warning instead of erroring out.

   This would make it easier to just grab the latest providers.json from https://oembed.com and use it. (Which could make Dynamically fetch oembed provider metadata (for url previews) #9877 possible)

Steps to reproduce

• add the providers.json from https://oembed.com/providers.json to your homeserver.yaml like so:

...
oembed:
  disable_default_providers: true
  additional_providers:
    - providers.json

• Start Synapse and immediately see the error that schemes are required

• After you have removed the entries which contain no scheme, start synapse again and you'll see the error with the spotify scheme.

Homeserver

Synapse Version

1.113.0

Installation Method

Docker (matrixdotorg/synapse)

Database

Workers

I don't know

Platform

Not relevant for the issue but I use Debian on a Dedicated Server and Synapse runs inside Docker

Configuration

No response

Relevant log output

# This is the log output for the first issue:

Error in configuration at 'oembed.additional_providers.<item 0>.<item 282>.endpoints.<item 0>':
  'schemes' is a required property

# This is the log output for the second issue:

Error in configuration at 'oembed.additional_providers.<item 0>':
  Unsupported oEmbed scheme (spotify) for pattern: spotify:*

Anything else that would be useful to know?

The docs should probably be refined a bit https://element-hq.github.io/synapse/latest/usage/configuration/config_documentation.html#oembed

I directly knew what the providers.json is and where to get it from, but I don't think that it is mentioned anywhere in the docs.
Also a direct link to the providers.json that gets shipped by default would be helpful. I at first thought Synapse wouldn't ship one contrary to what the docs state, but the problem was I only searched for oembed, and @clokep pointed me to the link where it can be found https://github.com/element-hq/synapse/blob/develop/synapse/res/providers.json

[bro2020]

I agree, this behavior should be corrected.