Apparently several years ago Twitter tried “monorepo” architecture… But on regular git. Because of their large codebase, and the fact that Git is not built for this at all, “git status” and “git log” would apparently take 5 minutes to run. They threw engineers at it for months but still Git was so slow. Forgot where I read this. Is this true? If this is the level of engineering at Twitter no wonder they have gone years without new features, they are waiting for git status to finish loading.
We solved those problems at Microsoft.
Microsoft did the same thing with windows source code. Faced the same issues but invented some new stuff like GVFS to make it work: https://devblogs.microsoft.com/bharry/the-largest-git-repo-on-the-planet/
So is FB the only one who buck-cracked it?
Meta (Facebook) also uses monorepo and was using git till 5-6 years back and offered to make it more efficient but Git foundation gave the same reason that mono repo is wrong way of doing things and refused the help that's why Facebook moved away from Git.
What did FB move to? Monorepo is simply wrong, though. You can't simply use git as if it's SVN.
What do you mean it's "simply wrong" Google has successfully used a monorepo since it's founding
You can the code base be large for a feed app with replies???
You’re right, the Salesforce SDK is probably bigger than Twitter’s codebase
https://gigamonkeys.com/flowers/ This is a nice read on Twitter's Engineering Productivity team. It has some bits on the monorepo challenges.
Amazon is mostly service base. I don't know of any monorepo here.
FB solved the problem. I love monorepo.
Is this googles logic to deprecating products? Once the product’s monorepo gets too slow to effectively work on it, they deprecate the product?
True story, though git usually completes under a minute now. If you know git internals join us!
If Elon is serious, but also keeps TC up, then I might just do that.