decisions that revision control systems have to make

copy from:

http://bazaar-vcs.org/RCSChoices

Here is a table of the decisions that revision control systems have to make. The three columns each represent something that every revision control system must decide upon.

• History - This is the aspect of a revision that is considered essential.

• Revision IDs - This is the way that the RCS internally names the revision for purposes of tracking, merges, etc.

• Rename Handling - How the RCS records file renames and apply them when doing updates and merges.

History	Revision IDs	Rename Handling
Snapshot	Hashes	Unique IDs
Changeset	Namespace	History tracking
Semantic Patch	Pseudo-random	None

Examples

Name	History	Revision IDs	Rename Handling
Bazaar-NG	Snapshot	Pseudo-random	Unique IDs
GnuArch	Changeset	Namespace	Unique IDs
Subversion	Snapshot	Namespace	History tracking
git	Snapshot	Hashes	File content
Darcs	Patch	Namespace	History tracking
quilt	Patch	Namespace	None
Mercurial	Snapshot	Hashes	History tracking

Matrix

The "History" axis is denoted in brackets after the name of the VCS.

Revision / Rename	None	History	Unique IDS
Namespace	quilt (patch)	Darcs (patch), SVN (snap)	Arch (cset)
Hashes	GIT (snap)	Mercurial (snap)
Pseudo-random			Bazaar-NG (snap)

Merge handling

Possibly this is a function of history, but another important decision is how to handle merges and branches. Eg monotone - commits to the same revision always create new branches