Check in with local and cloud vs simple git commit push

I was wondering why Unity Version Control manages commits and pushes this way. In a local repository, doing a ‘check-in’ acts as a local commit, requiring you to go to the Sync view (after you set up everything there, which is also not quite clear) to push your branches or check-ins. However, when connected directly to a cloud workspace (which is is still on you computer as local repo BTW), checking in commits and pushes at the same time. It feels very confusing coming from a Git background.

Hello @AndreaGalet!

What you ran into isn’t an inconsistency in check-in, it’s that Unity Version Control supports two different ways of working, and your two workspaces were each in a different one. Git only has one of them, which is why it feels strange coming from there.

Centralized (workspace connected directly to the cloud repo). There is no local repository at all. Your workspace talks straight to the server, and check-in writes the changeset directly there. It’s not “commit + push merged into one step”. There’s simply no intermediate repo to push from, so there’s nothing to sync. One operation and your work is on the server, visible to the team. This is the default and how most game teams use UVCS.

Distributed (a workspace working with a local repository). This is the Git-like mode: check-in commits to your local repo, and you push/pull changesets to the cloud through the “Sync to cloud” view. It’s there for working offline, or keeping work private until you’re ready to share it. It’s actually a bit more flexible than Git’s model because we support dehydrated clones and a very easy way to sync just a few branches.

So the mapping to your Git mental model is: distributed mode ≈ git commit then git push; centralized mode ≈ as if every commit were pushed instantly, because it happens directly on the server.

Why we’d suggest centralized for most projects: it’s simply fewer moving parts. One step instead of commit → push (→ and often pull/rebase when the push is rejected). You’re always working against the latest server state, so divergence stays small and merges stay rare. It also enables things a distributed model can’t give you: exclusive checkout (locking) so two people never edit the same binary asset, essential for scenes, prefabs and art, and it scales to huge repos because you never have to clone the full history to your machine.

And you have a fair point on the Sync view: the setup and the flow when you work distributed could be clearer, and that’s good feedback for us to improve on. But if your goal is just “commit and share my work”, the simplest fix today is to create your workspace directly against the cloud repository and let check-in do the whole job in one step.

2 Likes

Thank you for the answer, yep I understood how the system works and I was asking myself a couple of questions about QoL:

  1. I get why having the “Centralized model” could be better and effectively works better for artists and so on, but my question is… why wouldn’t the “Distributed model” work exactly as any git client?

For example: If the repository is a local one on my pc why couldn’t I have a check in and push directly on my Pending changes view instead of having to go to the Sync to Cloud view, set everything up once, then pulling and pushing from there? Wouldn’t it be better to have this flow in one window depending on the repo’s type? (Centralized or Distributed).

  1. A small, annoying thing for me with the “Distributed model” right now is that nothing notifies me whenever there is a new check-in on the cloud, why couldn’t I be notified about a new check in?

I feel like both of my questions are linked to how the “Sync to Cloud” works right now

As you can see here I set up everything I need but I don’t really understand the concept of the tree view with source and destination (on a side note: just write source and destination, I don’t like src and dst at all​:joy:).
Why should one be called source and the other one destination?