Hacker News Reader

Comment by Cloudef | Hacker News Reader

What's with the bunch of reddit crap on the front page recently?

https://old.reddit.com/r/OutOfTheLoop/comments/147fcdf/whats...

Why are we still posting links to claims they are "raising the price of access to their API from being free to a level that will kill every third party app" when they've made completely clear that over 90% of reddit apps won't have to pay a penny, and neither will the remaining 10% if they fix their inefficiency problems?

https://www.reddit.com/r/reddit/comments/145bram/addressing_...

> when they've made completely clear that over 90% of reddit apps won't have to pay a penny

Because the distribution of Reddit apps is not linear and if 90% of the long tail of Reddit apps (that nobody uses) are not affected that doesn't really change things.

Besides the difference between "all" and "10%", there's also the difference between "kill" and "require to fix gross inefficiencies".

These apps aren't grossly inefficient though, they've been substantially lower than the 84k per user per day API limit that reddit has held in the past for these applications, typically being below 500 requests per user per day. Even reddit's official app uses more API requests per day.

86,400 requests per day is Reddit’s historic per-user limit, but the number was picked as a maximum, not an average or a number any significant fraction of users would be hitting any significant fraction of the time. An app aggregating many users, if reasonably efficient, would be expected to have an average per-user rate over time much much lower than the single-user, single-moment absolute limit.

> not an average or a number any significant fraction of users would be hitting any significant fraction of the time

Mhm and no major third party reddit app is going anywhere near that with their 500 requests per day per user.

> An app aggregating many users, if reasonably efficient, would be expected to have an average per-user rate over time much much lower than the single-user, single-moment absolute limit.

The per user rate of these apps is 500 a day, which is reasonably efficient. Your point of them being "grossly inefficient" is just false.

Do you have a source for this claim other than spez himself, who has proven to be willing to, uh, let's say bend the truth?

Who would you consider a sufficiently authoritative source?

The developers themselves are every bit as authoritative on this topic while also not having been caught out lying. Given that, you'll forgive me if I choose not to put my chips with Reddit management.

I choose to trust the guy[0] who's been open in his communication and break down exactly why the new reddit API changes are forcing him to shut down his 3rd party app.

[0] https://web.archive.org/web/20230613060536/https://old.reddi...

Those two claims aren’t mutually exclusive.

Could you elaborate?

I'm not going to claim anything either way about the efficiency of these apps nor the potential (or lack thereof) to improve; but if they're shutting down rather than complying, this suggests that any efficiency improvements are not low hanging fruit.

Of course, now I'm thinking of that time a coworker insisted that $feature couldn't possibly be made more efficient and it was absolutely impossible to optimise it to take less than 20 minutes to complete, and I spent a few hours looking at what it was doing and got it down to 200 milliseconds…

…but I can't project random success anecdotes like that onto projects I'm not being paid to work on.

I’m interested specifically in Apollo, one of the most popular iOS Reddit apps, and the one that I use. Conveniently, its backend was recently open sourced [1] so how it uses the API is public.

Please cite one example of an “inefficiency problem” that Apollo could fix to put it on the path toward not “pay[ing] a penny.”

[1] https://github.com/christianselig/apollo-backend

Just at a glance, and maybe I'm reading it wrong, but this appears to be issuing six requests every five seconds:

https://github.com/christianselig/apollo-backend/blob/main/i...