The Past

Version control systems predating git are based on a client/server model. Clients are publishing changes on a central server and also downloading them from this central server.

• If the client can’t establish a connection to the server no code snapshot can be created

• Before creating a snapshot the local state has to be updated to include the latest changes from the server

• Only the server has all (meta) information about the code and it’s changes

        flowchart LR
    A(Client A) <--> S[(Server)]
    B(Client B) <--> S
    

This architecture has serious consequences:

• Not possible to create snapshots when being offline

• Without a connection to the server snapshots and their changes can’t be inspected

• Developers tend to publish their changes at the end of the work day

• Incomplete or even broken features are published and downloaded

• Code changes in single snapshots are rather big

• Big snapshots are error prone to conflicts

• If the server crashes and no backup of the server is available the whole history of snapshots is lost