phiresky

they are guaranteed to be sent in order and received in order so processing them in order should not be hard

on the receiving side the requests can still be processed fully sequentially, no need for parallelism. note this only solves the problem in this ticket and not the one you...

it does look like reqwest doesn't really have options for this, probably since http1.1 pipelining is mostly seen as superseeded by http2 multiplexing, which is better in most ways except...

Another alternative that's still simpler and less overhead than adding per-thread or per-community queues would be to add a field to each json activity or http request header with the...

it's a compatible change so it doesn't affect anything else

it just adds a header or json field that will be ignored if not understood. pseudo code: ``` post(/inbox).handle(activity): queues[activity.instance].add(priority=activity.sequence_id, activity) def queue_process(): queue = queues["lemmy.world"] last_id = 0 while...

but i have to say since completely ignoring activity sequence in a much much worse way than the thing above is how all of lemmy worked before 0.19 i don't...

i see what you're saying, though what i wrote now is now unrelated to http pipelining. imo we can do the simple solution and if it actually causes problems we...

splitting by community or post sounds neat but it's pretty complicated and sound like the kind of thing that's very hard to tune to balance the overhead, and the kind...

Try adding the dlls from the LeapSDK into the osvr-plugins-0 folder. Made it work for me in windows.