Eric Frohnhoefer may not have a job at Twitter anymore, but him fact checking dear leader constantly spewing nonsense about things he doesn't understand was a sweet moment.
Elon was talking about internal RPCs. Eric was talking about client side requests, as though that was a dunk. Eric was both wrong and attempted to insult his boss in public at the same time.
Take into account that between the two of them Eric is the expert.
It was Musk who failed to keep things professional. His first post was literally humiliating his own employees for writing bad code. It's his own company, he may not have been around, but it's his people now. Even if Twitter's Android client is not the best, playing the blame game is callous and petty. There is no point to trying to show that you are better than your employees either. It's the opposite of what a functional CEO should do.
It was callous and petty, and I can't imagine how frayed the nerves are of current employees. Almost every higher-up I've seen come into a company has done this to some degree, and it's always annoying, but to do it actually on the product to 100mm+ followers is a new level of cruel and spiteful.
"I was told ~1200 RPCs independently by several engineers at Twitter, which matches # of microservices. The ex-employee is wrong.
Same app in US takes ~2 secs to refresh (too long), but ~20 secs in India, due to bad batching/verbose comms. Actually useful data transferred is low."
There is no way to tell if this is right or wrong, we are not engineers at Twitter, and I suspect that no engineer there would dare to correct him. I'm pretty sure Twitter has an API gateway of some sort, it doesn't make sense that it's a batching issue.
But it doesn't matter who is right or wrong. The unprofessional atmosphere, the callousness, the one-upmanship is the problem. He is the CEO for God's sake. This is a PR nightmare. I know that it worked for Trump, but it isn't working even for him anymore.
I honestly don't understand what remote procedure calls have to do with the client. Surely the clients wouldn't interact in such a way with the microservices, there would be some sort of API gateway, maybe even one for each platform or service, like the backend for frontend pattern.
It's probably more complicated than this, and you can't exactly put it into a few words, but he tried, and it didn't make sense, but he still managed to call his employees idiots.
> An interesting thing about this claim is that not only is the implication wrong, Twitter probably has better evidence of its wrongness than any other company in its size class could have.
> There are very few companies that have a better distributed tracing setup w.r.t. getting actionable insights on the backend and the ones that have a better setup are much larger (Google, FB, etc.)
> Twitter client tracing also punches above its weight.
> Of course, the key people who did that work left or got laid off, but it's clear from the data that, if you're looking at why Twitter is so slow in, e.g., India, Uganda, etc., esp. on slow devices, tail latency comes from the network due to unreasonably large payloads + client.
> When leadership wanted to drive growth in India, they were into "visionary" stuff, e.g., "add features cricket watchers would like" when the app was unusably slow [on] a low-end device and the client performance team had been disbanded.