The first hour

SVN can be productively used within an hour. The mental model is simple: there is a server, you have a working copy, you update from the server, you commit to the server. Conflicts are explained linearly. New users from non-engineering backgrounds learn it without ceremony.

Git's first hour is harder. The staging area, the difference between commit and push, the meaning of HEAD, and the rebase/merge distinction all want explanation. Teams that throw new engineers into Git without scaffolding spend a month firefighting.

The architectural difference

Subversion is centralised: one server holds the canonical repository; clients hold working copies that are mostly snapshots. Without the server, you can edit files but not commit, log, branch, or merge.

Git is distributed: every clone is a complete repository. You can commit, branch, merge, view history, run blame, and rewrite locally without any network. Sharing happens via push and pull, but it is layered on top, not fundamental.

What we mean by performance

"Performance" in version control is multidimensional: clone/checkout time, log/blame speed, branch operations, working tree refresh, and concurrent contributor scaling. Git and SVN have very different profiles.

Clone versus checkout

SVN's checkout fetches only the latest revision of files - fast for shallow checkouts, no history pulled. Git's clone fetches the entire history by default - more bytes, but you have everything locally.

Two models, two costs

SVN branches are directory copies in the repository. The copy is cheap on the server (a metadata operation), but switching branches involves the network and a working-copy update. Git branches are pointers - a 41-byte ref. Creating, switching, and deleting branches is instantaneous and offline.

Two philosophies

Subversion (SVN), released in 2000, is a centralised version control system. The repository lives on a server; clients check out working copies, commit back to the server, and update from it. Git, released in 2005, is distributed: every clone is a full repository with complete history, and there is no architecturally privileged server.