How APIs die

Published by Beto Dealmeida on

Ramblings on the Twitter and Facebook APIs

Back in 2013 I wrote my own static site generator, which I still use, to run a blog. At the time I had already realized that blogs were no longer mainstream, so I wrote a blog engine that published my posts to the cloud. Each post would be announced in Facebook and Twitter, a small summary with a link to the blog post. People could comment on Twitter and Facebook, and periodically I would pull their replies and display them on the static post.

The Twitter API was already bad back then. I discovered there's no way to get replies to a given tweet. The workaround is to request all the tweets that mention you after the given tweet was posted, and scan each one to see if it is a reply to it. 8 years after, as far as I can tell, this is still the only way to do it.

The Facebook API used to be better. Initially I could post to Facebook from my blog engine, and collect replies from the platform. Eventually they made it more and more difficult. These days the only way to get it working is to create an application — but when you publish a post only people who have added that application can see your posts and reply to it. So it's practically useless.

Of course these platforms are not interested in people interacting with them from outside. If I could post programmatically to Facebook and fetch the comments and likes from my friends I would never visit the site (I already don't), and it would encourage them to also leave the platform. Once an API becomes popular enough, it dies.

Compare these two with the Mastodon API. Authentication is easy. Posting is easy. Fetching replies is easy. I miss the days when APIs were simple and everything had an RSS feed.