Post A1fplzwHoqJrGYAIoC by kaniini@pleroma.site
 (DIR) More posts by kaniini@pleroma.site
 (DIR) Post #A1fplzwHoqJrGYAIoC by kaniini@pleroma.site
       2020-11-29T00:38:00.969666Z
       
       1 likes, 0 repeats
       
       what I think I really want to do is fork Pleroma from before MFC got involved and took the technical direction in a direction that, frankly, wasn't good, and solve those problems in a more efficient way.
       
 (DIR) Post #A1fqckO0NJT6NSHc0G by libc@snack.social
       2020-11-29T00:44:48.075782Z
       
       0 likes, 0 repeats
       
       @kaniini i've been living under a rock in terms of what happened to pleroma development, what's the worst examples of the bloat?
       
 (DIR) Post #A1fqckhrBVLdN1ZSwi by kaniini@pleroma.site
       2020-11-29T00:47:34.254390Z
       
       0 likes, 0 repeats
       
       @libc modern Pleroma (1.0 and later) are designed more in the style of a rails app (using Oban, for example) than something that truly makes the most of BEAM.i want something that makes the most of BEAM and uses simpler queries to PostgreSQL instead of trying to do magic stuff in SQL.
       
 (DIR) Post #A1fqrSaGqTArVb2Hjs by kaniini@pleroma.site
       2020-11-29T00:50:21.149849Z
       
       0 likes, 0 repeats
       
       @libc there are also massive parallelization opportunities in ingress and delivery pipelines that can be achieved, simply by designing those pipelines to fully leverage the OTP ecosystem.much of the code added in the modern era has been largely imperative code that runs non-parallel (instead depending on Oban for parallelization).
       
 (DIR) Post #A1fr6lsrN4vHi421y4 by kaniini@pleroma.site
       2020-11-29T00:53:03.855216Z
       
       0 likes, 0 repeats
       
       @libc basically, on delivery to /inbox, the request worker should check the signature and if valid, stuff it in an inbound queue.  then you have an inbound-posts-scheduler worker which does a map-reduce every few seconds on the inbound posts queue.  this improves resilience while also improving multicore performance, as errors can just kick back to the queue.
       
 (DIR) Post #A1frCTFlYHzZfKSoBE by kaniini@pleroma.site
       2020-11-29T00:54:08.919257Z
       
       0 likes, 0 repeats
       
       @libc outbound, instead of trying to immediately deliver (or kick off to Oban), there should be an outbound queue, with recipient URIs and object IDs.  this should be stored in SQL, and again, map-reduced.
       
 (DIR) Post #A1fxvKd73mJXg19tFw by kaniini@pleroma.site
       2020-11-29T02:09:26.299914Z
       
       0 likes, 0 repeats
       
       @libc there is also something to be said about POSSE.  i want to be able to collide my social graphs together into a single view.  industrial standardization on ActivityStreams is meant to enable this, but Pleroma didn’t take it all the way.