[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: cvs or svn for common light weight users

From: Arthur Barrett
Subject: RE: cvs or svn for common light weight users
Date: Sun, 6 Jan 2008 23:13:15 +1100

> I wondered what the feeling was here about the future of cvs.
> And what a user with my needs should be getting into.

CVS is alive and well and being constantly developed.  It's been around
for more than 21 years and if it worked yesterday for you then it'll
work tomorrow just as well.

The 'ls' and 'dir' commands have been around for a while too and noone
seems to be saying that they need to be replaced in a hurry...  For the
use you described I would recommend sticking with the tool that already
is doing the job.

The project 'split' (for want of a better word) some time ago due to
different developers seeing different things as 'important': CVS, CVSNT,
OpenCVS, DCVS and SVN.  No tool is better than the other - they have
different strengths and weaknesses.  They are all actively developed and
all work on multiple platforms.

When choosing any software tool it is best to know what features you
require and then look for the tool that offers those features, or even
better look for what your goals are, then look for a process that
supports that, then look for tools that can implement that process.  For
instance knowing 'what' changed may be useless without knowing what else
changed - so you need to relate changes to one another and maybe
external events like project tasks, bugs or something else: so you would
need toos that support changesets and links to a system that tracks
those external events.

> CVSNT appears to remedy some of CVS's shortcomings, but does 
> not fix the
> overall brokenness of the CVS repository format (e.g., extremely slow
> tagging of large repositories).  It is also quite commercial 
> and rather
> out of the mainstream; for example, few CVS tools work with CVSNT.  If
> you switch to CVSNT you will make your eventual and 
> inevitable migration
> to a more modern VCS more painful.

I work on the CVSNT project (which is not a windows port - it runs on
Linux/Unix/Windows/Mac etc), so I'll dispell these myths.  CVSNT 3.x
uses a database backend and is currently in beta (and has been renamed
EVS", and CVSNT 2.5.x ( will continue
to use RCS for many years to come and has many advantages over database
backends for small projects like what you describe.

CVSNT ships with WinCVS and TortoiseCVS which are the most popular
windows CVS clients and has a download rate of around 1.4 million users
a year - it is the most common version control client used by commercial
software developers as far as we can tell.

Every CVS client also works with CVSNT - I know of not a single one that
doesn't and not a single one has ever been reported to the TortoiseCVS,
WinCVS or CVSNT newsgroups.  We actively work with Oracle and IBM
(Eclipse) and many other smaller developers who maintain CVS clients.

CVSNT project differs from the CVS project in that it has no 'charter'
governing what sort of changes will be accepted into the tree: this
means that people contibute a much wider variety of changes (some of
which 'traditional' CVS users find unnecessary): merge tracking, unicode
file merge, filename character set translation, rename, change sets,
inbuild ssh (client), disable protocols dynamically, failsafe audit,
commit id's, reserved, co-operative and unreserved edits etc etc  Some
of these have been 'backported' to CVS 1.12.x.  The majority of cases
where users require large numbers of tags can be replaced with other
features such as change sets.

Each 'CVS' has its own newsgroup - if you are interested in them then
ask on the individual newsgroups.  I suggest you ignore any posts that
tell you what is wrong with tool XYZ, look for posts that tell you what
is right about product ABC. 


Arthur Barrett

reply via email to

[Prev in Thread] Current Thread [Next in Thread]