This may be a dumb question,
But I am pulling back a table of ~500k rows, and hitting limits on the max web socket size.
Obviously the dirty answer is to increase that size, and the better answer is to change the view.
My question is, what is the negative to cranking that value up from 2mb to say 64mb?
More-and-less theoretical, in no particular order:
- CPU and memory consumption will increase on both sides
- Sending one massive payload means no other property sync messages can get through at the same time, leading to apparent property update delays
- DoS; maliciously sending lots and lots of very large payloads from the frontend to intentionally trigger the first two issues becomes a risk.
2 Likes
Good info.
Is there also a web-socket size limit for GAN? (And also a corresponding argument to control it)
I don't see one after a quick glance at the code, but I suspect it's less of an issue for GAN because generally the websocket is only used for 'control' messages (first party) and any expected 'heavy' transmissions happen out of band via regular HTTP.