Conversation

Notices

1. Christmas Personified as a Catgirl (moonman@shitposter.club)'s status on Monday, 07-Nov-2016 05:52:04 UTC Christmas Personified as a Catgirl

   • Annah
   @maiyannah i'm totally serious. if the api was strctured a bit differently it could be cached more effectively at the http level and even the php would get touched less often. and it wouln't even be a fundamental change, although it would not be trivial.
   
   when i'm building things for work i spend a lot of time thinking "how can i design this so it can be cached by varnish".
   
   for example, most people are camping out watching the public or TWKN timelines. those could be cached identically server side for everyone on a five to ten second cache expiration time, and then make a query for reply, star, repeat and follow notifcations to be pulled from a memory cache instead of returning them as part of the main streams or as a header on the api response. if the user switched to their follow list you could connect a websocket and server push only their data to them as it came in and never have to do a table scan every polling increment, you keep all their follows in memory because who cares its small.
   • Claes Wallin (韋嘉誠) likes this.
   • Claes Wallin (韋嘉誠) repeated this.
   • Christmas Personified as a Catgirl (moonman@shitposter.club)'s status on Monday, 07-Nov-2016 04:28:00 UTC Christmas Personified as a Catgirl
     best way to improve database performance is to avoid using a database where possible.
     Claes Wallin (韋嘉誠) repeated this.
   • Claes Wallin (韋嘉誠) (clackemovedtoheldscalla@quitter.se)'s status on Monday, 07-Nov-2016 05:06:01 UTC Claes Wallin (韋嘉誠)

     in reply to
     @moonman Yes. The greatest optimization is not to work more efficiently, but to simply do less.
   • Annah (maiyannah@community.highlandarrow.com)'s status on Monday, 07-Nov-2016 05:54:15 UTC Annah

     in reply to
     @moonman If you have specific ideas about how the API can be improved, PLEASE comment them in https://git.gnu.io/maiyannah/postActiv/issues/54 so they can be addressed when I redo the API, which will be my next major project after we've finished the postgres stuff and I've done the render code.  That might be a bit off and I don't want any of these ideas for improvement to get lost in the intervening time! :)
     Claes Wallin (韋嘉誠) repeated this.