I said i'd come back to this in a few months, and so i have. it has now been just over 1 year since i properly started releasing akkoma - the first release was 2022-07, so 2023-07 is as good a time as any.

as last time, i am going to go over some of what went well, and what didn't go so well, to reflect upon and maybe try and improve in the next 6 months.

also again, this is intensely personal so no bully

things that went well

adoption is still going strong

as twitter continues to die, we continue to prosper. server count is up to 530 at time of writing, comprising around 17,000 users. not bad i think. very scary though. very scary. it makes me nervous.

however, instance turnover seems to have settled down as instances stake their claim and people stop migrating every 5 picoseconds.

some more contributors appeared

not many, but a few. some PRs from people i'd never heard of before. which is something. takes some of the weight off, which is very nice. hopefully we can get a few properly involved and indoctrinated in the ways of AP so that everything doesn't just fall to like... me

the landing page is doing well on google

last time i did this i noted that getting the non-technical page up was a good thing - well, it's doing really well now; clicks to that homepage are up to 600 a month - though the technical dev page is on 2000. not sure how to best redirect people across, but it's 100% there in the results.

stability seems to be improving

IHBA is naturally rock-solid, but i have been reciving far fewer critical bug reports of late, which indicates that people aren't hitting the database "rot" issue that used to exist. whether this is due to improvements in query design or whatever else, i am unsure, but i won't complain.

things that did not go so well

extremely poor communication around security issues

when the media domain issue hit earlier in the year, nobody knew what was going on. nobody from any other project said anything, and we have no good ties with any of them, so i had to mitigate based on second and third hand information, and had to guess at details.

given that i'm pretty sure the shell of pleroma's team still hates me, and the other fork naturally does, i don't think i could have mitigated this at the time. and now that our one close ally in foundkey has more or less died, that leaves us exceptionally isolated.


the big sad stalling development

early in 2023 i hit another depressive episode which i'm only just sorta pulling myself out of the depths of. this massively cut into my motivation to be doing anything, let alone thankless fedi dev work. as such, i... just didn't do any. if it wasn't for a security issue in like may, i doubt i'd have looked at akkoma stuff at all in that time.

this is a really hard one to address since i struct out on my own, i am solely responsible for this project. but i must also make sure i'm ok in and of myself. i know it may be dissapointing for many people to see not a lot of activity on the project outside of maintenance, but... yeah i'm not quite sure what i'm supposed to say. sorry.

i think this ties into a wider theme of this project being far too broad for a single person to maintain alongside a fulltime job. it's just not feasible. i'm going to have to either project that this is a hobby project and will have lower expectations, or attempt to ignore the big sad. neither sound particularly good.

scope creep from pleroma days still causing issues

in the IRC we had a few people with very unusual setups, that proved really hard to assist in any meaningful way. this is due to the fact that pleroma and hence akkoma has a thousand configuration switches and potential setups that make it work any one of many different ways. this is hellish. just getting a user to coherently explain their setup and problem is tough enough, but when they could have toggled any one of 219831 switches into any configuration?

we lost a few potential instances due to this. we should reduce option bloat.

but also, not enough scope is also causing issues

when reddit decided to commit ritual suicide, the lack of support for its AP-enabled replacement caused many people to be upset. supporting it would increase the already horrifying size of transmogrifier, but not supporting it makes people sad.


people keep asking for more deployment methods, which i am really hesitant about doing. i sorta want to reduce deployment methods to a bare minimum due to how difficult it is to figure out issues on non-standard deployments...

support is still extremely difficult

there are too many users for me to manage. too many that are new to hosting software and need extra attention when providing support. too many that x-y-problem us in IRC.

i find it extremely difficult to get to even half of them, and have to hope someone else does (which does sometimes happen) - i'm not sure how to improve this in the slightest save for trying to recruit people.

things to improve by the next time I do one of these (november)

  • reduce the number of potential options to reduce the support workload. this WILL upset people. i will probably tell you in nov that it has been hard to deal with that.
  • update the FAQ section of the documentation with common support questions
  • ensure pleroma->akkoma migration is still easy and possible
  • don't treat myself too badly for having the sadde episode
  • continue managing the influx of users, try and provide targeted support to those that won't burn me out too fast (that is, don't help with standard sysadmin stuff as much)
  • create a proper way for people to report security issues?