CVS
http://www.cvshome.org/
AccuRev
http://www.accurev.com/
Aegis
http://aegis.sourceforge.net/
AllChange
http://www.intasoft.net/
Arch
http://gnuarch.org/
Arch Revision Control System
Bazaar
http://bazaar-vcs.org/
BitKeeper
http://www.bitkeeper.com/
ClearCase
http://www-306.ibm.com/software/awdtools/clearcase/
IBM Rational ClearCase
CM+
http://www.neuma.com/
Neuma
CMSynergy
http://www.telelogic.com/products/synergy/index.cfm
Telelogic
Co-Op
http://www.relisoft.com/co_op/
Code Co-Op
Darcs
http://abridgegame.org/darcs/
Fortress
http://sourcegear.com/fortress/
Fossil
http://www.fossil-scm.org/
Git
http://git.or.cz/
LibreSource Synchronizer
http://dev.libresource.org/
Mercurial
http://www.selenic.com/mercurial/
Monotone
http://monotone.ca/
OpenCM
http://www.opencm.org/
Perforce
http://www.perforce.com/
PureCM
http://www.purecm.com/
SourceAnywhere
http://www.dynamsoft.com/Products/SAWStandalone_Overview.aspx
DynamSoft
Subversion
http://subversion.apache.org/
Superversion
http://www.superversion.org/
Surround SCM
http://www.seapine.com/surroundscm.html/
svk
http://svk.elixus.org/
Team Foundation Server
http://msdn2.microsoft.com/tfs2008/default.aspx
Microsoft Corp.
Vault
http://sourcegear.com/vault/
Vesta
http://www.vestasys.org/
Visual SourceSafe
http://msdn.microsoft.com/ssafe/
Microsoft Corp.
$Id$
Version Control System Comparison
This is a comparison of version-control systems. It is split
into several categories and sub-categories under which the
systems are checked.
Repository Operations
Atomic Commits
Support for atomic commits means that if an
operation on the repository is interrupted
in the middle, the repository will not be
left in an inconsistent state. Are the
check-in operations atomic, or can
interrupting an operation leave the
repository in an intermediate state?
No. CVS commits are not atomic.
Yes. Commits are atomic
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Commits are atomic.
Commits are atomic.
Commits are atomic.
Commits are atomic.
Commits are atomic.
Yes (but need to verify)
Yes.
Yes.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
No. VSS commits are not atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits (checkins) are atomic.
Yes. Commits and updates are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Yes. Commits are atomic.
Files and Directories Moves or Renames
Does the system support moving a file or directory to
a different location while still retaining the history
of the file? Note: also see the next section
about intelligent merging of renamed paths.
No. Renames are not supported and a manual one
may break history in two.
Yes. Renames of both files and directories are supported.
Supports controlling of symbolic links as well.
Yes. Renames are supported.
Yes. Renames are supported.
Yes. Both moves and renames are supported, while
maintaining history.
No. Renames are not supported.
Renames are supported, moves can be accomplished similarly
to SourceSafe.
Yes. Renames are supported.
Yes. Renames are supported.
Yes. Renames are supported for files and directories.
Yes. Renames are supported.
Yes. Renames are supported.
Yes. Renames are supported.
Yes. Renames are supported.
Yes. Renames are supported.
Yes. Renames are supported
Yes, supported
since version 2011.1.
Yes. File and folder renames and moves are directly
supported.
Yes. The unit of checkout/checkin is a directory
tree. Files and directories can be added,
deleted, and renamed between versions.
Renames of files are supported.
Renaming a directory requires creating a new one,
moving the files and deleting the old one.
Moved file histories are preserved.
Affects the whole history, it's like renaming a
file in the CVS repository. There is a kludgy workaround
using "share-rename,move,delete" that gets what you
want.
Yes. Renames are supported.
Yes. Directories are first-class controlled entities
in ClearCase. Even supports controlling of
symbolic/hard links.
Yes. Renames and move are supported but the working copy
needs to be up-to-date before doing a rename/move
operation. This operation will be committed directly.
Yes. Both moves and renames are supported, while
maintaining history.
Renames are supported for most practical
purposes. Git even detects renames when a file has been
changed afterwards the rename. However, due to a peculiar
repository structure, renames are not recorded
explicitly, and Git has to deduce them (which works well
in practice).
Yes. Both moves and renames are supported,
while maintaining history.
Yes. Both moves and renames are supported, while
maintaining history.
Yes. Both moves and renames are supported,
while maintaining history.
Moves and renames are supported. History is retained.
Intelligent Merging after Moves or Renames
If the system keeps tracks of renames, does it support
intelligent merging of the files in the history after
the rename? (For example, changing a file in a renamed
directory, and trying to merge it).
No. Renames are not supported at all, much less
intelligent ones.
Unknown. FILL IN.
Partial. Move is implemented as copy+delete/obsolete, full
revision history is maintained across copying, so changing
a file in the copied directory does the right thing.
No. "svn help mv" says "Note: this subcommand is
equivalent to
a 'copy' and 'delete'." There's a
bug report about it.
Yes. Renames are Intelligent.
No. Renames are not supported.
Yes. Renames can be merged intelligently.
No. Same as Subversion.
Yes. Renames can be merged intelligently.
Yes. Renames are intelligent.
Yes, renames are intelligent but should be explicitly
specified using "darcs mv" because they are not detected
automatically.
Unknown. Probably Yes.
Unknown. FILL IN.
Yes, intelligent merging after renames is supported. the
Mercurial book says:
"If I modify a file, and you rename it to a new name, and
then we merge our respective changes, my modifications to
the file under its original name will be propagated into
the file under its new name. (This is something you might
expect to 'simply work,' but not all revision control
systems actually do this.)"
Yes. Intelligent merging is fully supported.
Unknown.
See point above.
Yes, intelligent renames are supported.
Unknown. FILL IN.
Unknown. FILL IN.
No, renames are not intelligent.
Unknown. FILL IN.
Yes, renames and moves are intelligent in ClearCase and
are handled well. Every unique file is given a UUID and
directory entries map names to UUIDs, so moving a file
around preserves its UUID. Doesn't work across VOBs.
Yes. Renames are intelligent. However, the rename should
be made by the system, in order to be detected in the right
manner.
Unknown. FILL IN.
No. As detailed in the Git
FAQ:
"Git has a rename command git mv, but that is just a
convenience. The effect is indistinguishable from removing
the file and adding another with different name and the
same content."
Unknown. FILL IN.
Yes, intelligent renames are supported.
Yes, intelligent renames are supported.
Yes, intelligent renames are supported.
File and Directory Copies
Does the version control system support copying
files or directories to a different location at the
repository level, while retaining the history?
No. Copies are not supported.
Copying is supported through symbolic links
(but all linked files are treated as the same file
version). Moves are fully supported with the history
retained.
Yes. Copies are supported.
Yes. And it's a very cheap operation (O(1)) that
is also utilized for branching.
Yes. An inexpensive operation that can be used for
sharing files in multiple places. On deploy, you have
the option of deploying only one of the shared files or
all of them.
No. Copies are not supported.
No. Copies are not supported.
Yes. Same as subversion.
No. Copies of files and directory structures are
not supported.
No. Copies are not supported.
No. Copies of files and directory structures are
not supported.
Yes. Copies are supported.
No. Copies are not supported.
Yes. Copies are supported
No. "Tracked" copies are currently not supported, i.e.
monotone does not track information about copied files.
No. Copies are not supported.
Yes, fully supported with full
history.
Yes. Copies are supported.
Yes. A new package/branch can be based on any
existing version without affecting the past
history. (This is also an O(1) operation.)
Copying doesn't retain history, moving does.
Yes. Copies are supported up to a point.
Yes, and it's a very cheap operation (update the target
directory to include the new file/directory).
Yes, through use of hard links. (But some
limitations in Windows environments)
No, copies will start their own history.
Copying doesn't retain history, moving does.
No. Copies are not supported.
Yes - you can create a branch. But the GUI has no option to
view the old history. The
Power-Tool tfpt has the option /followbranches to show
the history of the file branch's ancestors
No, copies are supported, but will start their own
history.
No, copies are supported, but will start their own
history.
Copying is not directly supported. Duplicating
the file and adding it to the tracking would not copy
the original file's history.
Remote Repository Replication
Does the system support cloning a remote repository to get
a functionally equivalent copy in the local system? That
should be done without any special access to the remote
server except for normal repository access.
That used to be the case indirectly, by using
CVSup
by John Polstra (which requires running the cvsupd
daemon on the server). However, cvsupd has become
unmaintained, now has problems with CVS 1.12.13, and is
written in Modula-3, which makes it non-recommended.
Indirectly, either by using the
svnsync
tool which is part of the Subversion distribution on
recent versions, or by using Chia-liang Kao's
SVN::Mirror
add-on or Shlomi Fish's
SVN-Pusher
utility.
Yes. CM+MultiSite can be configured to clone a repository
so that it continues to act as a single repository. Options
include cloning only from the main site (i.e. not allowing
updates from the clone) and restricting the set of files
transferred to a cloned site.
Yes.
No.
Yes.
Yes, using the proxy server.
Yes.
Yes.
Yes.
Yes.
Yes.
Yes.
Yes.
Yes.
No.
Yes. Via the Perforce Proxy (P4P) tool,
and full
repository replication.
Yes. Using the PureCM Proxy Server.
Yes. Replication is a fundamental part
of the design.
Repositories are always replicated on local machines.
There is no central server.
Not directly possible with the included GUI or
command line tools; ssarc and ssrestor might be usable
Yes, as long as you have the (more expensive) Distributed
package.
Not really applicable for ClearCase, but see next point.
Yes, but is not documented and its based on the dataflow
feature of the LibreSource Synchronizer.
Not directly possible with the included GUI or command
line tools; Some SQL Server tool might be usable.
Yes. Using the "git clone" command.
TFS Proxy is available but the replica isn't an
equivalent copy.
No.
No.
Yes. Entire repository can be downloaded without special
permission (the copy will not track upstream changes).
More traditional
cloning
requires authorization (this is typically provided to
'anonymous' users).
Propagating Changes to Parent Repositories
Can the system propagate changes from one repository to
another?
No.
With AccuReplica, the replica server has all the meta-data
and fetches file data as needed by replica users; all
write operations pass automatically from the replica to
the master server.
No.
Yes, using either Chia-liang Kao's SVN::Mirror
script or the svn-push utility by Shlomi Fish.
Yes. In CM+MultiSite, changes made at the slave are, by
default, propagated to the Main(master) library, as well
as to all other Clones (slaves). You may also propagate
changes between unrelated repositories containing some of
the same source.
No.
Yes.
Yes.
Yes.
Yes.
Yes.
Yes.
Yes.
Yes.
Yes.
No.
Yes, via
remote depots.
No.
Yes.
It's a peer-to-peer system, which keeps all replicas of
the repository in sync.
Not directly possible with the included GUI or
command line tools; ssarc and ssrestor might be usable
Yes, as long as you have the (more expensive) Distributed
package.
Yes, using ClearCase Multisite.
Yes, it's what we call a dataflow.
Not directly possible with the included GUI or command
line tools; Some SQL Server tool might be usable.
Yes, it is possible.
No.
No
No
Yes.
Repository Permissions
Is it possible to define permissions on access to different
parts of a remote repository? Or is access open for all?
Limited. "pre-commit hook scripts" can be used to
implement various permissions systems.
Yes. Access can be defined per stream (branch) using
access control lists.
Yes. Permissions may be defined based on role and
area of the repository
Yes. It is possible to define permissions on access to
different parts of a remote repository based on the
permission systems of the underlying protocol.
Basic access control can be implemented through a
contributed hook script. ACL support for the
Bazaar server is planned.
No.
Yes. Aegis relies on the UNIX permissions system to
implement permissions for files in the repository.
FILL IN
Yes. The WebDAV-based service supports defining HTTP
permissions for various directories of the repository.
Yes. Permissions can be defined at all levels of the
system.
Yes. Permissions are defined by data, primarily,
not by location. If location is a part of the data, it may
be used to define permissions by location. Permissions may
apply to a branch, file, problem report, test case, etc.
Access may be extended based on peer group, manager, and
access lists.
No.
Same as subversion.
Yes. It is possible to lock down repositories,
subdirectories, or files using hooks.
Current monotone has specific read permissions with which
you can control the access to certain parts (aka branch
patterns) for known user ids, but less specific write
permissions. Basically, if a user is allowed to write to
a server, they can write everything there pretty
much unrestricted.
Permissions are defined on a per-branch
basis.
Yes. (more than half a dozen of permission levels that can
be set in a file by file basis)
Yes. Permissions can be set against repositories, streams
(branches/labels), folders and files using Access Control
Lists.
Yes. Access permissions for each package (the
unit of checkout/checkin) can be different.
Access permissions for a branch can be different
from the basis package.
First access (joining the project)
requires administrator's approval.
Subsequent access to that project is not controlled.
Project specific permissions (read, write, delete, destroy)
can be set per user; but see "Networking Support":
this makes "Repository Permissions" a hindrance to
accidental damage but cannot prevent intentional damage.
No, though a single server can serve many repositories.
Yes, a Unix-like permissions model is used, which maps
onto Windows domain-based authentication in
multi-platform environments.
Permissions are set for the whole repository or branch.
Yes. SourceAnywhere Server Manager can define access to a
repository per user or group and user access rights to a
project.
See contrib/hooks/update-paranoid that ships with Git. See
the path_rules code for the closest equivalent to svnperms.
Yes. You get set permissions for each team project, folder,
file.
Yes. Revisions can be set, and overridden, and the
repository, project and folder level.
Yes. Revisions can be set, and overridden, and the
repository, project and folder level.
Permissions are set for the whole repository.
Changesets' Support
Does the repository support changesets? Changesets are a way
to group a number of modifications that are relevant to each
other in one atomic package, that can be cancelled or
propagated as needed.
No. Changes are file-specific.
Partial support. There are implicit
changeset that are generated on each commit.
Yes. Changesets (called changellists) as
well as labels are fully supported.
Yes. Change packages are known as updates. By default, an
update is required to make any change. The update may be
checked-in, differenced, promoted, retrieved, propagated,
yanked (i.e. removed from history), etc. each in a single
operation. Baseline alignment is performed based on the
status (i.e. promotion level) of the update. Updates also
record changes to directory structure: move, add, remove.
Yes, AccuRev provides robust functionality for change
sets (called change packages in AccuRev) including viewing
differences by change packages and merging changes from
stream to stream by change package.
Partial. Changes may be associated with Change Requests
and promoted /baselined/released as a set.
Partial support. Changes are grouped into changesets,
but cannot be cancelled individually yet.
Same as subversion.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Yes. Changesets are supported.
Not exactly. Vesta uses a related concept of
configurations instead, which some has similar
properties.
Yes. Changesets are the default.
No. Changes are file-specific.
Yes. Changesets (or tasks) are fundamental
to the way Synergy works.
Not supported in this way. Extensive branching
support gives similar benefits. (eg each changeset
can be given a branch). Also optional UCM feature
gives something like this (each changeset is a
"stream").
Partial support. There are implicit changeset that are
generated on each commit.
Not exactly. SourceAnywhere uses a related concept of
configurations instead, which some has similar
properties.
Yes, Changesets are supported, and there's some
flexibility in creating them.
Yes. Changesets are the only possibility.
Yes. Changesets are supported.
Yes. Changesets are supported.
Not directly. Changeset information is available in
"manifest" files but no commands are provided to
automate changeset operations.
Tracking Line-wise File History
Does the version control system have an option to track the
history of the file line-by-line? I.e., can it show for each
line at which revision it was most recently changed, and by
whom?
Yes. cvs annotate
Yes. Available from both the GUI and CLI.
Yes
Yes. (svn blame)
No.
Yes. View revision tags.
No.
Yes. (svk blame)
Yes. (bk annotate)
Not in the command line client, but ViewARCH,
a web-interface for Arch, has it.
Yes. (bzr annotate).
Yes. (darcs annotate)
Yes. (hg annotate)
Yes. (mtn annotate)
Yes. aeannotate
Unknown. Probably not.
Yes, an annotation feature is present.
Yes, annotation is available through the GUI.
No, but it would be easy to implement a tool that
did this, as the Vesta repository provides direct
filesystem access to all versions.
Not directly, but it's possible to compare
any two versions using a visual differ.
Not directly, but it's possible to compare
any two versions using a visual differ.
Probably, if you're a sufficiently proficient hacker with
their scripting language.
Yes, "cleartool annotate"
Yes, locally without any server connection with the
standard graphical Java client.
Yes. (SAW annotate)
Yes. (git blame).
Yes. (tf annotate).
Yes. Both standard Blame and Line History (Blame on
selected sections of a file) are supported.
Yes. Both standard Blame and Line History (Blame on
selected sections of a file) are supported.
Yes. This capability is provided with the
'fossil annotate'
command as well as through the
built-in web interface.
Features
Ability to Work only on One Directory of the Repository
Can the version control system checkout only one directory of
the repository? Or restrict the check-ins to only one
directory?
Yes.
Yes. AccuRev provides functionality to define
feature streams in which only the subset of code is seen.
A group of developers can then be restricted to work only
from that stream so they are only allowed to check in
changes to that subset of code.
Yes. Any arbitrary set of files or directories may be
checked out or in.
Yes.
Yes.
Yes. Any arbitrary set can be checked out
and worked on. Similarly, arbitrary restrictions may be
applied for check-in, including file ownership.
No.
Yes.
No. All changes are made repository-wide.
It is possible to commit only a certain directory.
However, one must check out the entire repository as a
whole.
For checkouts: No. For checkins: Yes.
It is possible to commit only a certain directory.
However, one must check out the entire repository as a
whole.
No. All changes are made repository-wide.
It is possible to commit changes only in a subset of the
tree. There are plans for partial checkouts.
It is possible to commit changes only in a subset of the
tree. However, one must extract the entire tree to work
on it.
No. All changes are made to a project as
a unit
Yes. Changes to a sub-directory of the repository
are supported.
Yes.
Yes and no. The unit of checkout/checkin (called a
package) is a directory tree. Most projects use
more than one. Once created, a package must be
checked out/in as a unit.
No. All changes are made to a project as
a unit, but it's possible to access each file's
history separately.
Yes.
Yes and no. Files and directories are checked out and in
individually, however you have to work in the context of a
project, which consists of one or more directories.
Yes, using snapshot view load rules.
It is possible to commit only a certain directory.
However, one must check out the entire repository as a
whole.
Yes. SourceAnywhere can define the user access right to
each project and users can be restricted to work only on
the projects they have check out/in right.
No. However, commits could be restricted somewhat, see
the "Repository Permissions" item.
Yes.
Yes.
Yes.
No. Checkouts are of the entire repository, however,
commits can be limited to any subset of files (e.g., just
the contents of a subdirectory).
Tracking Uncommited Changes
Does the software have an ability to track the changes in the
working copy that were not yet committed to the repository?
Yes. Using cvs diff
Yes. The functionality is available through both the GUI
and the command line interface.
Yes. Using the comparison facilities.
Yes. Using svn diff.
Yes. Use Updates | Delta | Delta Update. Or
right click a file or directory and do a compare to
workspace.
Yes. Local changes are detected and shown immediately.
Changes can be collected in a local buffer before being
committed to the repository.
Yes. Using svk diff.
Yes. Using "bk diffs".
Yes, using "tla changes".
Yes, using "bzr diff".
Yes, using the repository difference.
Yes, using "darcs whatsnew".
Yes. Using aediff.
Yes. Using hg diff.
Yes. In a similar fashion to CVS.
Yes. Using cm diff
Yes. Using aediff.
Yes.
Yes.
Yes. Intermediate immutable snapshots can be
taken during an active checkout (with vadvance).
These intermediate versions can be treated just
like checked in versions: they can be replicated
to other repositories and used as the basis for
branches.
Yes, using built-in visual differ/editor.
Yes, using the integrated diff tool.
Yes, either using the integrated diff tool or
a user-configured external diff tool.
Yes, "cleartool diff".
Yes, with the Synchronizer Studio (default Java client) or
with the standard diff command (diff -r .
.so6/xxx/REFCOPY/)
Yes. Using saw diff.
Yes. Also, branches are very lightweight in Git,
and could be considered a kind of storage for
"uncommitted" code in some workflows. Also see the
"git stash" command.
Yes. Using tf diff or "Pending Changes" in Visual Studio.
Yes. Using DiffMerge.
Yes. Using DiffMerge.
Yes. Using
'fossil diff' or 'fossil gdiff'.
Tracking status of all files is also available.
Per-File Commit Messages
Does the system have a way to assign a per-file commit message
to the changeset, as well as a per-changeset message?
No. Commit messages are per change.
No. Commit messages are per change.
No. Commit messages are per check in.
No. There is no such feature.
Yes. Out of the box CM+ is configured to
prompt for messages (i.e. comments) only per change.
However, the schema is pre-configured so that you may
prompt on a per file basis as well (typically done at
checkout time as the entire change is normally checked
in with a single operation.
Yes.
Yes.
No. There is no such feature.
Yes. It is possible to have a per-file
commit message.
No.
With respect to pure Bazaar: No. At least one
plugin (bzr-gtk) supports it though.
No.
No.
No. Commit messages are per-changeset.
Unknown.
No. Commit messages are per change.
No. Commit messages are per change.
Not exactly. The unit of checkin is a directory,
and commit messages are assigned at that level,
not to individual files. Since configurations are
also versioned, they also have commit messages.
No. Commit messages are per change.
They go to all project members and update
their repositories.
Since changesets are not supported, yes.
Yes.
Yes, assuming a comment on the branch is sufficient
for a per-changeset message.
No. Commit messages are per changeset.
No. There is no such feature.
No. Commit messages are per changeset.
No. Commit messages are per changeset.
No. Commit messages are per-changeset.
No. Commit messages are per-changeset.
No. Commit messages are per check in.
Technical Status
Documentation
How well is the system documented? How easy is it to
get started using it?
Excellent. There are many online tutorials and
resources and an online book. The command line client
also provides an online comprehensive help system.
Excellent. There is a full set of documentation available
in PDF format available at
AccuRev
Documentation as well as context-sensitive help
in the GUI.
Excellent. There is a full set of documentation available
as both help and PDF.
Excellent. There is a full set of documentation available
as
both help and PDF.
Very good. There is a free online book and some online
tutorials and resources. The book is written in
DocBook/XML and so is convertible to many different
formats. The command-line client also provides a good
online help system that can be used as a reference.
Very good. There is a self-demo/tutorial
to get you started quickly. Administration is minimal.
So normal developer use requires only a 1 to 2 hour
training session (or equivalent guide) to introduce you
to concepts and capabilities (e.g. like updates, options).
Customization documentation is also extensive but should
normally be accompanied by a 2-day to 4-day course for
GUI, Process, Data and Application set customization.
Fairly poor. There are two tutorials, but there is no
reference. Installing and getting started with the GUI
is very easy though.
Relatively poor, but improving. There's
a work-in-progress
book as well as
the Wiki and some
external Articles and Tutorials.
Medium. The documentation is given in several large scope
troff documents, that are only usable as not-so-PDFish
PDF documents, and as text documents that lack any
formatting. It is very hard to get started using
it with the online resources. The content is of good
quality, but otherwise not made very accessible.
Medium. There are two online tutorials and a
comprehensive online documentation. The command line
client also supplies a reference page. However, some of
the documentation is out of date or incomplete.
Excellent. Apart from online help in the command
line client there exist tutorials, a reference
card ("Quick Start Guide"), several full fledged
guides and references, and documents on
specialized topics, such as migration from other
VCS systems and different workflows. The
documentation comes in html and plain-text formats. The
API of the underlying library is fully documented.
In the UI design of the command line client
special attention was paid to make it easy to get
started with Bazaar.
Good. The manual contains a brief tutorial and a solid
reference. Every sub-command can print its usage.
Because the command-set is small and the model is
simple, many users find it easy to get started.
Very good. There is a comprehensive help at the BitKeeper
site. Each command is documented in its own man page,
and the client contains a help tool that offers
an integrated help system.
Very good. There is a
companion
book and a wiki. Every command has integrated help.
Good. There's a lot of documentation available in PDF
and HTML formats. The client supplies documentation
for every command.
Well documented.
Very Good (HTML and command line help).
Very Good (HTML and command line help).
Quite thoroughly (HTML, man pages, published
papers, a book-length research report).
Very good. Step-by-step tutorial and HTML help
is included.
Medium. Help file which is sometimes useful.
However, the interface is reasonably intuitive so
documentation isn't needed as much.
Medium. Lots of books, plus somewhat clunky set of HTML
pages, but has some radical concepts which can cause real
problems really quickly. They recommend a day's training
for basic users, more for more advanced users. Took a
while to become fluent.
Extensive online help in Windows Help / UNIX man page
format, also PDF-based documentation. However the
complexity of the tool can mean a lengthy ramp-up
time.
Medium. There are an online tutorial and some comprehensive
online documentation. Installing and getting started with
the GUI is very easy though.
(update/commit-next-next-next-finished)
Good. There's an overview and tutorial on the web site,
and integrated help for every command.
Good. There's an online help for every command detailing
all the flags. The man pages are extensive, but tended to
be confusing (possibly improved by now). There are many
tutorials, blog posts and some books (some of which are
online).
Good. A comprehensive documentation in the MSDN Library.
Many Step-by-Step tutorial videos online.
Good. All features are documented, plus a getting-started
guide.
Good. All features are documented, plus a getting-started
guide.
Fossil is well documented. Man page-type help is available
for all commands using 'fossil help'. Fossil also includes a
built-in web interface
permitting easy visual exploration and administration.
There are also
tutorials,
user guides,
and
a wiki
available on the Web.
Ease of Deployment
How easy is it to deploy the software? What are
the dependencies and how can they be satisfied?
Good. Out of being the de-facto standard,
CVS is available on most systems and is easy
to deploy.
Excellent. All that is required is to download the
binaries for the appropriate platform and run the
installer. The installation package is self-contained. No
additional software is needed. AccuRev supports most UNIX,
Linux, and Windows platforms and deploying AccuRev to a
multi-platform environment is straight-forward.
Good. Various out-of-the-box configurations supplied
including ITIL support. These may be tailored or used as
is. Time to deploy depends on how much configuring is
done.
Excellent. All that is required is to download the
binaries for the appropriate platform and run the
installer. The installation package is self-contained.
No additional software is needed. Surround SCM supports
Windows, Mac OS X, Linux, and Solaris platforms and
deploying Surround SCM to a multi-platform environment is
straight-forward.
Excellent. An arch service is nothing but a
filesystem-space hosted by any of its supported
protocols (FTP, SFTP, WebDAV, etc.). The arch client
is written in C, and is portable across UNIX systems
(and on Win32 only with a UNIX emulation layer).
Very easy. Bazaar has an installer for MS Windows
and packages for some major Linux distributions,
FreeBSD, and Mac OS X. The dependencies for
manual installation are listed on the Bazaar
website.
Very good. darcs requires few external libraries,
however you need the Glasgow Haskell Compiler if you
cannot find a binary. To start working, just "darcs
init".
Good. All that is required is downloading a binary
for the system and installing it using the installation
script. The package is self-contained and is easy to
set up.
The Aegis binary should be installed as SUID-root, and
so requires root privileges to install. It also not very
portable to Win32 systems. Other than that, Aegis supports
an easy autoconf or RPM/apt-based installation process.
A Subversion service requires installing an Apache 2
module (if one wishes to use HTTP as the underlying
protocol) or its own proprietary server. The client
requires only the Subversion-specific logic and the
Neon WebDAV library (for HTTP). Installation of the
components is quite straightforward, but will require
some work, assuming Subversion does not come prepackaged
for one's system.
Yes. Typical installation need only be done
on the server (with a single shortcut established on the
client). This assumes file system connectivity. For IP
only connectivity, installation is also required on
remote clients. Installation is typically a couple of
minutes. No dependencies unless web interface is used,
in which case an Apache server is required. A download
is available from Neuma's web site and takes you right
into a self-guided fully working demo version.
If Java 1.4 is installed, deployment of Superversion
usually takes two clicks.
In addition to installing subversion, users are required
to install the subversion perl bindings and a few modules
from CPAN.
Excellent. Binary packages are available for all
popular platforms. Building from source requires
only Python 2.3 (or later) and a C compiler.
Excellent. It is possible to copy or compile the executable
to the user's machine, without any configuration or
external dependencies.
Very good. Install the RPM, or build from tarball and
install the init script.
Very good. Perforce is very easy to deploy.
Very good. PureCM is very easy to deploy.
Medium to Good. There is a detailed installation guide
for setting it up using a binary kit. RPMs and
Debian packages have been recently released.
There are no dependencies on other software. There is a
bootstrap package available to build Vesta from using
"make".
Very easy to deploy, since there is no central
server. Can be configured to use e-mail or LAN (or both) for
synchronization. For e-mail, requires MAPI-compliant
e-mail client.
Very good - an installation package which does the work.
When you create a repository it installs the executables in
a directory and you can run them from there if you need to.
Medium. There is a detailed install guide for setting it
up using a binary kit and a set of scripts. However
it still took several tries to get it properly installed
and configured. The Windows client has a slightly clunky
Windows installer.
Poor. ClearCase is very difficult to install in
general. At least, setup for a new site is quite
complex. Installing additional servers (eg
repository servers) is less so.
Excellent. It is managed by JavaWebStart with links on any
LibreSource repository web page.
(links: create workspace, update, commit, studio...)
Excellent. DynamSoft SourceAnywhere is extremely easy to
install. It is totally written in C++ from scratch, which
means that you don't need any additional components and
frameworks to support the installation.
Good. Binary packages are available
for modern platforms. C compiler and Perl are
required. Requires cygwin on Windows, and has some
UNIXisms.
Installation is quite complex. Needs IIS, MS-SQL Server
and Reporting Services. A own installation guide with
step-by-step guide is available. Allows separating in
data and app tier.
Easy. Prerequisites are IIS 5 or higher, SQL 2000 or
higher. Install takes minutes.
Easy. Prerequisites are IIS 5 or higher, SQL 2000 or
higher. Install takes minutes.
Fossil is a single, stand-alone executable file that
can be installed anywhere in the user's execution path.
Precompiled binaries are available for GNU/Linux, Windows,
and Mac OS X. Fossil only has the ZLIB compression tools
library as its only dependency.
Command Set
What is the command set? How compatible is it with
the commands of CVS (the current open source defacto
standard)?
A simple command set that includes three most commonly
used commands (cvs commit, cvs update and cvs checkout)
and several others.
Very extensive but not compatible with cvs.
Very extensive but not compatible with cvs.
Very extensive but not compatible with cvs.
A CVS-like command set which is easy to get used to
for CVS-users.
CM+ has several dozen commands that can
be used both for operation and configuration of the product.
As CM+ is change-based, commands are substantially
different than CVS. The GUI is used primarily and implemented
on top of the command set. As well, CM+ covers a full ALM
suite and can be extended beyond, so there are many more
generic commands for browsing, reporting, etc.
There is little need to memorize a command set because
all actions take place in a GUI. A part of the terminology
used in the application is borrowed from CVS.
A CVS-like command set which is easy to get used to
for CVS-users.
A CVS-like command set with some easy-to-get-used-to
complications due to its different way of work and
philosophy.
A complex command set that involves many operations
just to get started. Not CVS-compatible. (albeit
support for such basic operations was contemplated)
Note that Aegis is a Software Configuration Management
system and not just a simple version control system,
which may justify this extra complexity.
Many commands are compatible with CVS or BitKeeper. However,
there are many other commands for it for different uses.
Aliasing of commands is possible so it it may be possible
to make it more compatible.
Tries to follow CVS conventions, but deviates
where there is a different design.
The command set is fairly compact and the core commands
are easy to understand. Follows CVS in a few places,
but since the model is different most commands are
unique.
Tries to follow CVS conventions, but deviates where there
is a different design.
Tries to follow CVS conventions, but deviates where there
is a different design.
A CVS-like command set that is familiar to existing CVS
users.
Very extensive but not compatible with CVS.
A CVS-like command set which is easy to get used to
for CVS-users.
The command set is unrelated to CVS. Most of the
time, users use about 5 commands. Few ever need
to know more than about 20 commands.
Basic commands are compatible with CVS.
A bit of an afterthought. It's possible to do basic
things, but it's really geared up for using the GUI.
An extensive and powerful command set,
which has some CVS similarity, though the architecture
is so different that it quickly moves away for anything
but the basics.
Excellent. All tools are available through the
command-line. Not very compatible with CVS though.
Basic commands available (commit/update), but it's really
simple to use the GUI. Ant task are also available.
Very extensive but not compatible with CVS.
Command set is very feature-rich, and not compatible with
CVS.
The command set allows more operations than the GUI
but isn't compatible with CVS.
Extensive command set via command line and GUI. Not
compatible with CVS.
Extensive command set via command line and GUI. Not
compatible with CVS.
Basic command set with most core commands identical
to CVS (though the option switches are often different).
Networking Support
How good is the networking integration of the system?
How compliant is it with existing protocols and
infra-structure?
Good. CVS uses a proprietary protocol with various
variations for its client/server protocol. This protocol
can be tunnelled over an SSH-connection to support
encryption.
Good. (proprietary protocol using TCP/IP)
Good. Uses TCP/IP and HTTP/HTTPS
Excellent. Arch can utilize a multitude of protocols
for its service, which is nothing but a dumb remote
filesystem server. Currently supported protocols include
FTP, SFTP, WebDAV (remote file access over HTTP),
as well as any remote filesystem protocol (NFS, SMB).
Excellent. Works natively over HTTP (read-only),
FTP and SFTP without having Bazaar installed at
the remote end. Works over HTTP, SSH and a custom
protocol when talking to a remote Bazaar
server. Supports RSYNC and WebDAV (experimental)
through plugins.
Good. (proprietary protocol using TCP/IP)
Good. Darcs supports getting patches over HTTP, and
getting and sending patches over SSH and email.
Very good. The Subversion service can use either
WebDAV+DeltaV (which is HTTP or HTTPS based) as its
underlying protocol, or its own proprietary protocol
that can be channeled over an SSH connection.
Very good. File system connectivity,
TCP/IP connectivity and Web connectivity may be
intermixed. MultiSite connectivity is over TCP/IP,
as is License server. Works well with SSH, NFS, SMB, etc.
Good. Network support based on RMI is integrated
seamlessly. Encryption and HTTP tunnelling are planned
for the near future.
Very good. svk uses SVN::Mirror to retrieve remote
repository. There has been plans to add VCP support
to SVN::Mirror so it will be able to mirror from arbitrary
remote version control systems.
Good. Repositories can be checked out from remote
over HTTP, and BitKeeper also sports its own proprietary
protocol for communicating between one repository and
the other.
Poor. Aegis is filesystem-oriented and so can be networked
only via NFS (network file-system) or a similar protocol.
There exists some HTTP-functionality, but it is quite
limited.
Excellent. Uses HTTP or ssh. Remote access also
works safely without locks over read-only network
filesystems.
Good. Uses a custom protocol called "netsync".
Good. Uses its own proprietary client/server protocol.
Good. (single TCP/IP socket)
Good. (single TCP/IP socket)
Networking is inherent to the system. The
repository exports both an NFS interface and an
RPC interface. The checkout and checkin tools
automatically contact a remote repository when
required to perform an operation.
Uses the simplest LAN interface:
copying files between shared directories.
VSS uses a Windows network share which has to be writeable
for the VSS users (since this means doubling maintenance
for new users). Add user in VSS and to share permissions.
the share is most often world-writeable, as is the default
when creating a share) It does not perform well over a
slow network connection.
Good (single TCP/IP socket)
Poor. Uses an *extremely* chatty RPC protocol for
most ClearCase operations, plus NFS or SMB for
accessing the files themselves. Typically servers
should be deployed locally (i.e.: on the same LAN) as
the client workstations for acceptable performance.
Good. Use of HTTP to get through firewalls.
Good. (single TCP/IP socket)
Excellent. Can use the native Git protocol,
but also works over rsync, ssh, HTTP and HTTPS.
Good. Use of HTTP(S).
Good. HTTP and HTTPS only.
Good. HTTP and HTTPS only.
Excellent. Fossil integrates both server and client into
a single application. A
built-in web server
permits graphical administration and navigation over
HTTP and HTTPS; as well as providing a
bug ticketing system
and a
simple wiki
for documentation.
Portability
How portable is the version-control system to various
operating systems, computer architectures, and other
types of systems?
Good. Client works on UNIX, Windows and Mac OS.
Server works on UNIXes and on Windows with a UNIX
emulation layer.
Excellent. The server runs on most UNIX, Linux
and Windows platforms. The client runs on all of these
platforms and on Mac OS X.
Available only for windows platforms.
Excellent. Clients and Servers work on UNIX,
Windows and Mac OS X.
Excellent. The server runs on Windows, Mac OS X, Linux
and Solaris platforms. The client runs on all of these
platforms.
Good. Clients and Servers work on Unix, Linux, and
Windows. MAC OS X port pending. Moving server from one
platform to another is a copy operation only. Can have
different platforms for different servers in a MultiSite
configuration. Easily configurable Web client also
supported. No CR/LF issues. Scripts are all portable as
well.
Excellent. Clients and servers work on any Java
1.4-compatible platform. There is official support
for Windows, Linux and OS/2.
Good. Clients requires subversion and its perl bindings.
Very good. Binaries are available for most common UNIX
systems and for Windows 98 and above.
Medium. The source is portable across all UNIXes,
but the Windows version work only using cygwin, and even
then not entirely natively.
Good. The source is portable across all UNIXes,
but requires a UNIX emulation layer on Windows. (need to
verify). A service can be hosted on any platform
that sports a suitable Internet service.
Works on MS Windows, Linux, Mac OS X, FreeBSD,
UNIX, and basically on any system that has a
recent Python port. With case-insensitive file
systems there are some issues that can be avoided
by using a graphical front-end. On MS Windows
there is a plugin to support tracking of symbolic links
even if they are not supported natively by the
file system.
Very good. Supports many UNIXes, Mac OS X, and Windows,
and is written in a portable language.
Excellent. Runs on all platforms supported by
Python. Repositories are portable across CPU
architectures and endian conventions.
Excellent. Executable is portable across all UNIXes and
Win32.
Good. Portable across all UNIX systems.
Excellent. Runs on UNIX, Mac OS, BeOS and Windows.
Excellent. Client and Server run on Windows, Linux,
Solaris and other UNIXes. The client also runs on Mac
OS X.
Good. It should be portable to any UNIX system.
Currently it runs on Digital/Compaq/HP Tru64 UNIX
and Linux on several different CPU architectures.
Ports to Solaris and FreeBSD are planned but
haven't begun yet.
Windows only: starting with Win95.
The Microsoft Product is Windows only.
MainSoft
ships a version of it for some UNIX platforms.
Very good - various flavours of Unix,
Windows (only NT family for the server), VMS, and
possibly other systems.
Medium. Available on Windows, and several selected
flavours of UNIX (not including any other Linux
distribution than some versions of Red Hat Enterprise
Linux, SUSE Linux Enterprise Desktop and Ubuntu Linux).
Excellent. Clients and servers work on any Java
1.5-compatible platform. (Windows, Linux and Mac OS X )
Good. The server runs on Windows only. Clients can work on
any platform that SWT (Standard Widget Toolkit) supports,
including Windows, Linux, Mac, Solaris, AIX, HP-UX, SCO
Unix, FreeBSD and so on.
The client works on most UNIXes, and there's a native
MS-Windows build. The cygwin build seems to be workable as
well.
The Server and Client needs Windows. A third party company,
Teamprise, has developed a client for Eclipse, which means
Linux, Mac and other UNIXes support. The Project SvnBridge
allows access using SVN clients but needs to run on
Windows.
The server, and standalone client, are Windows only. The
Eclipse plugin is cross-platform, as is the command-line
client.
The server, and standalone client, are Windows only. The
Eclipse plugin is cross-platform, as is the command-line
client.
Fossil integrates both server and client into
a single application which can run on any POSIX-like
operating system (e.g., GNU/Linux, Mac OS X, MS Windows,
et al).
User Interfaces
Web Interface
Does the system have a WWW-based interface that can be
used to browse the tree and the various revisions of the
files, perform arbitrary diffs, etc?
Yes.
CVSweb,
ViewVC,
Chora,
and wwCVS.
No.
Yes.
Yes. Its own built-in web-interface.
Yes.
ViewVC,
SVN::Web,
WebSVN,
ViewSVN,
mod_svn_view,
Chora,
Trac,
SVN::RaWeb::Light,
SVN
Browser,
Insurrection
and perl_svn.
Aside
from that, the Subversion Apache service provides a
rudimentary web-interface.
Yes. Can be configured to restrict which operations
are allowed by which users, so that customers may access their
requests without seeing development team data.
No.
Yes. Same as Subversion.
There's ViewARCH,
and ArchZoom
which are works in progress.
Yes, using WebDAV.
Yes, several:
Loggerhead,
Webserve,
Bzrweb,
and
Trac.
darcs.cgi
is included in the distribution.
Yes.
Yes. The web interface is a bundled component.
Yes. ViewMTN
and a Trac plug-in.
No.
Yes, P4Web.
Yes.
Yes: Vestaweb.
Since this functionality is always
available locally, there is no need for web interface.
It is possible to code one using the API, but no official
or third-party one exists.
Possibly.
Yes. Web views are supported.
Yes, without diff features but with a better awareness
support. (allow to know at any time on each version each
one is working on)
Currently not.
Yes. Gitweb is included in the distribution and there's
also cgit.
Web Access is available as download for free.
Yes.
Yes.
Yes. Fossil also includes a web-based
bug ticketing system
and
built-in wiki.
Availability of Graphical User-Interfaces.
What is the availability of graphical user-interfaces for
the system? How many GUI clients are present for it?
Very good. There are many available GUIs:
WinCVS, Cervisia (for KDE),
TortoiseCVS (Windows Explorer plug-in).
A single, comprehensive,
Java-based GUI is provided. The GUI has the same
look-and-feel on all platforms.
A single windows based interface is provided.
Good. BitKeeper ships with several
GUIs for performing common tasks. I'm not aware
of any third-part GUIs.
Very good. There are many available
GUIs: RapidSVN (cross-platform),
TortoiseSVN (Windows Explorer plug-in), Jsvn (Java), etc.
Most of them are still under development.
A complete native GUI is available for all platforms.
Excellent. Windows and Unix/Linux GUI as well as web GUI.
Extensively configurable via simple menu files, browser
files, etc. Can customize the set of to-do lists by
user/role, same for menus, pop-up menus, default visible
tabbed reports, etc. GUI also used for all admin and for
process and data schema customization. Also plug-in for
Visual Studio, Eclipse, etc. and File Browser (Windows).
A GUI is integrated.
No GUIs are available.
There are
tlator,
Octopy,
and ArchWay
and possibly others under development.
There are several graphical front-ends in development, see
the Bazaar
Plugins page and the
Third-party Tools page. Notable are QBzr (Qt) and bzr-gtk
(GTK+), which can be considered beta quality.
Work is also being done on integrating Bazaar with Windows
Explorer, Eclipse, Nautilus, and Meld.
None to speak of. (There is a modest graphical
interface to a few commands in the distribution, but it
is not being developed currently.)
There is tkaegis.
History viewing available with hgit extension;
check-in extension (hgct) makes committing easier.
TortoiseHg
provides a Windows shell extension, and GNOME/Nautilus
integration. Some third-party IDEs and GUI tools (e.g.
eric3, meld) have integrated Mercurial support.
Yes, there is
mtn-browse
(which should be considered the "best of breed"),
monotone-viz
and guitone.
There's a
complete
list of tools available on the Monotone wiki.
No GUIs are available.
Yes, P4Win, P4V and others based on the available libp4
library. There are also plugins for Eclipse, Visual
Studio and other environments.
Cross-platform GUI for Windows, Linux, Mac OS X
and other UNIXes.
No GUIs are available, but the repository has a
C++ API, and it is not hard to write one. (At
least three different project-specific ones have
been written by users at Compaq and Intel.)
The system is GUI-based by design.
Standalone GUI comes with it, plus SCCI plug-in for
MS Visual Developer Studio. There is an Eclipse
plug-in.
A couple of GUIs. A motif-based one
(even on Windows) allows most functionality but is clunky.
A nicer Java one allows developer work but not much
administrative stuff. Has an SCCI plug-in, though it
doesn't handle network problems well.
Supplied for both Windows and UNIX. GUI tools are
typically not as solid as the command-line tools
though.
One written in Java/SWING and available on any OS that
is automatically launched from the repository web page and
another one which is an Eclipse plugin.
The system is GUI-based by design.
Gitk is included in the distribution. Qgit and Git-gui tools are also available.
TFS client integrates into Visual Studio.
Standalone Windows GUI, Visual Studio integration, and
cross-platform Eclipse integration.
Standalone Windows GUI, Visual Studio integration, and
cross-platform Eclipse integration.
Aside from the built-in web-interface, there is
Fuel, which is
a cross-platform and open-source GUI for Fossil, written
using the Qt toolkit.
License
What are the licensing terms for the software?
GNU GPL (open source)
Proprietary, named-user licensing.
Proprietary, named-user and concurrent licensing available.
GNU GPL (open source)
GNU GPL (open source)
GNU GPL (open source)
Proprietary, binary only license. Pay per use license,
with an option for a costless license for developers of
open source software. Used to have a gratis, downloadable
license, which was intended for the development of open
source software. It had a
problematic license,
and was discontinued starting at April 2005.
Proprietary, named and floating licensing.
GNU GPL (open source)
Apache/BSD-style license. (open source)
Network licenses and user licenses. No minimum checkout time,
and automatic license checkin on idle. License server included
in product. Professional and Enterprise editions. Enterprise
includes customizations, additional applications, and full
multiple site capability. One Server license per site. Total
license cost per user typically less than $1000 + 18% annual
mtce.
GNU GPL (open source)
Perl License. (open source)
GNU GPL (open source)
GNU GPL (open source)
GNU GPL (open source), but moving soon to
BSD or CPL (also open source).
A proprietary, binary only, commercial license.
Price
starting at $800 per seat for the first year and then
a $160 for continuing support for the subsequent years. The
latter payment is optional and required only for support,
as the product can be used without it. Free for
open source projects (no support in this case), and there
is also a ‘free for 20 users’ offer.
A proprietary, binary only, commercial license.
Price
starting at $1000 for 5 users
GNU LGPL (open source)
Proprietary, short text key. 30-day
full-featured trial. Free to "observers"
(members who don't make changes).
$159 per workstation.
VSS Ships with MSDN, and can also be purchased
standalone or with other tools.
Prices negotiable with salesman.
Server is typically roughly 20,000 British Pounds.
Clients are 4,000 British Pounds. Per-year costs of 18%
of original.
Proprietary, with floating license supported. License
server contacted for each ClearCase operation, which
obtains a license to be used for a next duration (8 hours
by default - most people lower it to 2 hours or less,
while 30 minutes is the minimum). Prices are several $k per
license plus a yearly maintenance fee. Typically 1-3 users per
license required, depending on activity. Multisite requires
additional licensing.
QPL - The Qt Public License (open source)
Proprietary, named-user licensing.
GNU GPL v2 (open source).
Commercial license.
Commercial, per-user with no separate server license.
Commercial, per-user with no separate server license.
Simplified BSD license,
also known as the BSD 2-clause license (open source).