Matrix Synapse module which implements MSC3886 Simple HTTP rendezvous server

This is an implementation of MSC3886 for Synapse.

Usage

1. Install the module to make it available to your Synapse environment:

   pip install matrix-http-rendezvous-synapse
   

2. Enable the module in your homeserver.yaml:

   modules:
     - module: matrix_http_rendezvous_synapse.SynapseRendezvousModule
       config:
         prefix: /_synapse/client/org.matrix.msc3886/rendezvous
   

3. Make the module available at the actual API endpoint in the Client-Server API by adding this to your homeserver.yaml:

   experimental_features:
     msc3886_endpoint: /_synapse/client/org.matrix.msc3886/rendezvous
   

4. Enable additional CORS headers for the API endpoints within the listeners section of your homeserver.yaml:

   listeners:
     - type: http
       experimental_cors_msc3886: True
       # ... rest of the HTTP listener config
   

5. Run Synapse with the asyncio reactor enabled:

   SYNAPSE_ASYNC_IO_REACTOR=1 python -m synapse.app.homeserver
   

Configuration options

Apart from the prefix the following config options are available:

• ttl: The time-to-live of the rendezvous session. Defaults to 60s.
• max_bytes: The maximum number of bytes that can be sent in a single request. Defaults to 4096 bytes.
• max_entries: The maximum number of entries to keep. Defaults to 10 000.

An example configuration setting these and a custom prefix would like:

modules:
  - module: matrix_http_rendezvous_synapse.SynapseRendezvousModule
    config:
      prefix: /rendezvous
      ttl: 15s
      max_bytes: 10KiB
      max_entries: 50000

experimental_features:
  msc3886_endpoint: /rendezvous # this should match above

listeners:
  - type: http
    experimental_cors_msc3886: True
    # ... rest of the HTTP listener config

Memory usage

max_entries and max_bytes allow to tune how much memory the module may take. There is a constant overhead of approximately 1KiB per entry, so with the default config (max_bytes = 4KiB, max_entries = 10000), the maximum theorical memory footprint of the module is (4KiB + ~1KiB) * 10000 ~= 50MiB.