5 ms·

So, are there 1,000 RPCs needed to render the timeline or not? The Android engineer didn't seem to know the difference between an RPC and a request from an end

So, are there 1,000 RPCs needed to render the timeline or not? The Android engineer didn't seem to know the difference between an RPC and a request from an end user client. The former timeline engineer disabled public responses to his tweet and has refused to elaborate. Neither of them backed up their contradiction with a rebuttal. Edit: Was curious so did some digging and found more: > @elonmusk: > 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. > Part of today will be turning off the “microservices” bloatware. Less than 20% are actually needed for Twitter to work! > And we will finally stop adding what device a tweet was written on (waste of screen space & compute) below every tweet. Literally no one even knows why we did that …

salad-fan 4y ago
Did you even read the Twitter thread? The number (~20) was provided right there.
- programmarchy 4y ago
  Yes, and I also know the difference between a request from an end user client and a server-to-server RPC which was referenced in Musk’s tweet. The Android engineer is confusing the two, and so are you apparently.
  - salad-fan 4y ago
    Space Karen didn’t make that distinction in his original tweet: https://twitter.com/elonmusk/status/1591853644944932865 https://twitter.com/elonmusk/status/1591853644944932865 In fact by referring to the “App” making “RPCs” (2nd sentence), the most straight forward interpretation is that he was talking about client -> server calls.
    - programmarchy 4y ago
zero_shift 4y ago
That depends on proximate versus ultimate causes. Does the Android app "need" the 20 initial HTTP requests or does it "need" the thousands of services that ultimately keep that data in good shape? I think most people, talking specifically about a client program, mean the former, as the latter is out of its control
BulgarianIdiot 4y ago
Elon Musk said "the App makes >1000 RPC". Calling the backend services the "App" would be atypical. So it's expected than an App engineer (the one on Android, the most popular Twitter App) would assume Elon is talking about calls made by the App, wouldn't he? There may be 1000+ microservices fed the firehose and so on on the backend. Sure. But... those microservices also existed two weeks ago, when Twitter was running just fine.

salad-fan 4y ago

Did you even read the Twitter thread? The number (~20) was provided right there.

programmarchy 4y ago
Yes, and I also know the difference between a request from an end user client and a server-to-server RPC which was referenced in Musk’s tweet. The Android engineer is confusing the two, and so are you apparently.
- salad-fan 4y ago
  Space Karen didn’t make that distinction in his original tweet: https://twitter.com/elonmusk/status/1591853644944932865 https://twitter.com/elonmusk/status/1591853644944932865 In fact by referring to the “App” making “RPCs” (2nd sentence), the most straight forward interpretation is that he was talking about client -> server calls.
  - programmarchy 4y ago

zero_shift 4y ago

That depends on proximate versus ultimate causes. Does the Android app "need" the 20 initial HTTP requests or does it "need" the thousands of services that ultimately keep that data in good shape? I think most people, talking specifically about a client program, mean the former, as the latter is out of its control

BulgarianIdiot 4y ago

Elon Musk said "the App makes >1000 RPC". Calling the backend services the "App" would be atypical. So it's expected than an App engineer (the one on Android, the most popular Twitter App) would assume Elon is talking about calls made by the App, wouldn't he? There may be 1000+ microservices fed the firehose and so on on the backend. Sure. But... those microservices also existed two weeks ago, when Twitter was running just fine.