Hacker News Reader

Comment by jiggawatts | Hacker News Reader

Stop repeating this, because Elon had the gist of it correct, if not the specifics.

a) Twitter is atrociously slow for something that displays a mere kilobyte of text at a time.

b) Much of that is because of hundreds of client-server JavaScript requests on first load. Not technically RPCs, but just as slow and looking nearly identical on the network.

c) Most of the rest is because of a thousand microservices chatting away at each other. Not technically client-to-server, but still taking significant time.

Elon accidentally conflated the network path of (b) with the performance issues caused by (c).

They're both bad! They both need fixing! The fact that Musk mixed up doesn't matter.

View on HN

> Much of that is because of hundreds of client-server JavaScript requests on first load.

Isn't this not true though? Based on the responses to his Tweet only one GraphQL request is made which batches the whole thing.

> Most of the rest is because of a thousand microservices chatting away at each other. Not technically client-to-server, but still taking significant time.

But what does this have to do with India? If the slowness is due to calls between microservices hosted on (from my understanding) their own hardware, why would it matter if the initial GraphQL request comes in from India, since all the rest of the backend requests would be the same?

> his Tweet only one GraphQL request is made which batches the whole thing.

Literally bang in the middle of an F12 network trace showing dozens of requests, mostly JavaScript. Not JSON RPC, per se, but still an overhead taking up seconds, especially on low-end Android devices.

> But what does this have to do with India?

Everything. Nitter uses 40KB of code & data to display the same content that takes Twitter megabytes.

People in subsequent threads have pointed to the gigabytes of "data" usage by the Twitter Android app, which is just absurd for an app that displays primarily text.

It's a bloated pig of an app that performs poorly on 5G networks in the United States, and hence terribly on slow and overloaded networks separated from Twitter's data centres by an ocean.

"It works fine for me" says the engineer connected to the WiFi in Twitter HQ.

> Literally bang in the middle of an F12 network trace showing dozens of requests, mostly JavaScript. Not JSON RPC, per se, but still an overhead taking up seconds, especially on low-end Android devices.

I'm pretty sure that's my point? How are you equating "dozens" to "1200" (or even "hundreds" in your last comment) as though Elon is in the right ballpark?

As for the rest, I don't know what your point is, where did Elon say anything in that thread about how much data they were sending? He's quite literally responding to someone saying it was the amount of JavaScript being downloaded and telling them they're wrong, it's the "1200 RPC calls". I'm pretty sure the point he was attempting to make is that it's slow due to thousands of serial "RPC" calls made between the app and Twitter (which will have a higher RRT to India) but as you've demonstrated that's not what it's doing.

I'm not saying Twitter is well designed, I'm just not pretending like Elon somehow had the right idea when you've had to stretch it this far. If he had a proper technical understanding then we wouldn't have to have such a discussion of what he "actually meant".

That's correct, I am working in mobile world since 2005 and the latency of mobile networks is the most bad thing when you build mobile apps. The more requests you do the more latency you introduce to the user, especially if it's fetching data for UI, even if you do it asynchronously you have to wait for the data to update the UI.

> Much of that is because of hundreds of client-server JavaScript requests on first load. Not technically RPCs, but just as slow and looking nearly identical on the network.

This is not true at all. I have a look and it's not that there are many REST requests it's that the requests send back a load of data. For Example the timeline sends back a 1MB JSON response. It's heavily gzipped so comes in at 100KB over the network but it takes quite a while for the whole request.

In places like india the internet is slow and many people are using VPNs. I emulated this and I can see that over VPN on a slow internet connection it's like 10 seconds to load the home page.

He was right on the details of the two different issues, because those were explained to him.

But he conflated two things that he didn't really understand and reached a completely wrong conclusion. He's shared that conclusion with confidence and never admitted he was wrong. That's a fatal flaw for someone who holds a Chief Engineer title at a rocket company.

He's a CEO who is has been incredibly successful at selling a vision and getting people to do what he wants. He is not an engineer. He should stop role playing as one.

A good engineer (!) would have been able to understand the difference, just saying.

Reply on news.ycombinator.com