On the client side, there is usually a local connection pool. When a burst of traffic comes in, the client needs to either wait for the pool to free up or establish a new connection, which adds latency. This latency hit wouldn’t occur with multiplexing.

With multiplexing, systems like pgbouncer would be unnecessary.

Also, even with a thread-per-connection, you can still quickly exhaust the servers resources when you have lots of connections because threads have a lot of overhead. Reducing the number of connections needed would greatly increase the number of clients that a database can serve.

> You may have heard that if you have an MTHFR variant, you should avoid folic acid and should take other types of folate, such as 5-MTHF. However, this is not true. People with an MTHFR gene variant can process all types of folate, including folic acid. Folic acid is the only type of folate shown to help prevent neural tube defects (NTDs).

> A qualified health claim is supported by less scientific evidence than an authorized health claim. FDA requires that qualified claims be accompanied by a disclaimer that explains the level of the scientific evidence supporting the relationship.

> Unlike authorized health claims, FDA does not issue regulations for qualified health claims.

That does seem to indicate supplement manufacturers have broad latitude to make claims provided they are disclosed as such.

We were expanding our product to a new language that used non-ASCII code points. Part of the system involved invoking binaries using text as input.

Locally, everything worked great. Once deployed, we got corrupted text output. As soon as we SSH’d on to the server to inspect, everything started working again.

It turns out that SSH servers can modify the LANG environment variable. The default value on our servers didn’t support Unicode, but LANG was updated as soon as we connected via ssh. It was a head scratcher for sure.

The product was social-event focused (classes, festivals, etc.) so we focused on multiplayer audio-video experiences rather than general purpose browsing.

One of my favorite memories was when someone used our collaborative YouTube playback to set up a karaoke room. WebRTC added a little latency, but it was close enough to work.

> The effort required to adapt them has dropped

AI is an entirely different situation because the effort required to copy has dropped by multiple orders of magnitude. You used to be able to build in the open without worrying about copycats because the vast majority of people didn’t want to spend the effort. Now (with AI), even someone with the slightest, most fleeting whim can copy your work.

It’s great that you’re open to being adapted. There’s nothing wrong with that. But if you’re not open to having your ideas outright taken, then it’s not safe to build in the open any longer.

Take this game, for example: https://news.ycombinator.com/item?id=47698455

Within an hour, someone had cloned the game with addition mechanics that multiple people mentioned they like more: https://news.ycombinator.com/item?id=47729573