Conversation

Notices

1. Embed this notice
   Una (unascribed@sleeping.town)'s status on Monday, 28-Nov-2022 08:53:37 JST Una

   great protocol we've got here >.>

   • Embed this notice
     Una (unascribed@sleeping.town)'s status on Monday, 28-Nov-2022 08:53:33 JST Una

     in reply to

     this is no longer a joke https://jort.link/

     hypolite likes this.
   • Embed this notice
     Una (unascribed@sleeping.town)'s status on Monday, 28-Nov-2022 08:53:34 JST Una

     in reply to

     brb starting a Jortage LinkShield® service that requests and caches the linked URL and serves that cache to fedi server UAs, and redirects everyone else to the original URL

     hypolite repeated this.
   • Embed this notice
     Una (unascribed@sleeping.town)'s status on Monday, 28-Nov-2022 08:53:36 JST Una

     in reply to

     by boosting any post in this thread you exacerbate this problem. every boost causes like a dozen new instances to request all of my bio links

   • Embed this notice
     Una (unascribed@sleeping.town)'s status on Monday, 28-Nov-2022 08:53:36 JST Una

     in reply to

     well yes, any moderate to large server, or god forbid, a fleet of servers (such as a CDN or load balancer/etc) can absorb such a load with no issue — I certainly have no issues with the load imposed by such a request flood

     but especially as surplus metal becomes cheaper and cheaper (you can get a dedicated server from OVH for $6/mo, not even to mention how little you can get a VPS for now, or running a server on your home connection with a raspberry pi) there are a lot of sites out there with no real brunt that are nonetheless expected to be able to handle loads comparable to large organizations

     mastodon is supposed to be a form of champion of the indie/decentralized web, but its design instead abuses and punishes those that create it

     in this very thread I've had someone talking about their experiences having their wiki that's linked in their bio constantly taken offline just from people replying to them or boosting their posts

     not even to mention the effect this can have if someone links an expensive endpoint. for example I run a rendering service, which up until now didn't have much in the way of caching/etc because it was a fairly tough problem to design a solution for, and to finally implement

     if someone were to link that old version of the service on mastodon with even a modest following, it would go offline immediately, and not recover for an hour from the backlog — what do you think happens if the request fails?

     if someone wanted to keep that service offline for any reason, they would only have to post a link to it, and keep editing that post every hour, or making new posts every hour. the backlog would only grow, and grow. after enough time, instances do give up, but you only need to put a new job in the top of the queue...