Add param to enable ws per-message deflate #365
Conversation
Tornado has its own per-message deflate compression option, which compresses each WebSocket message. The compression level should be roughly equivalent to PNG compression, depending on whether the message is JSON or binary (CBOR). The encoding/decoding time will be much faster than protocol PNG compression. This param should be enabled when wire size is important, e.g. not connecting to localhost.
| if not cls.use_compression: | ||
| return None | ||
|
|
||
| return {} |
There was a problem hiding this comment.
Do we want to document the behaviour of this? At least a copy paste from Tornado's docs can help:
If this method returns None (the default), compression will be disabled. If it returns a dict (even an empty one), it will be enabled.
Should we expose the compression and memory level as well? IMO at least a comment describing the defaults would suffice.
There was a problem hiding this comment.
I didn't see any of the websocket server's params documented anywhere, I'm happy to add a note if you can point me at a place to leave it. I assume the Tornado docs will be the first point of reference for anybody looking at this.
Compression and memory level only became available in Tornado 4.5, so if we exposed them it might take extra work to support kinetic. I didn't feel like adjusting those settings would be particularly valuable.
Tornado has its own per-message deflate compression option, which
compresses each WebSocket message. The compression level should be
roughly equivalent to PNG compression, depending on whether the message is
JSON or binary (CBOR). The encoding/decoding time will be much faster
than protocol PNG compression.
This param should be enabled when wire size is important, e.g. not
connecting to localhost.