|
Message
From: Richard Tierney<rt-opencores@c...>
Date: Tue Apr 18 10:50:15 CEST 2006
Subject: [oc] kindly convert opencores cvs repository to subversion (svn)
Jeremy Fillingim wrote:
> I understand that we are walking on thin
> ice with this discussion, and that everyone has their preferred work flow.
Indeed. :)
I've been having this argument for years. I won't repeat your points,
but, briefly:
1 - Why are commercial tools lock-modify-unlock, rather than
copy-modify-merge? Can you name a proper commercial tool that uses
copy-modify-merge? Why is this?
2 - Imagine you're a project manager with say, 20 developers, each of
greatly differing experience and ability, and each working on 100 source
files. How do you set this up? I'll tell you how I would set it up: each
developer owns their own source files. They're responsible for those
files. The other developers cannot get arbitrary access to those files.
A developer carries out the requried changes to their own files, except
in exceptional circumstances, such as code reviews, holidays, and so on.
3 - the whole concept of merging different changes into a software
source file is just bizarre. The whole file should ideally be about 400
lines long, and certainly no longer than 1000 lines. One person needs to
understand the entire file intimately. How could you, as the project
manager, possibly allow two different developers to simultaneously
modify two different parts (or, in the worst case, the *same* part) of
the same file?
I've worked in one commercial environment where we used CVS. This worked
because we agreed on the rules above, so merging was not an issue. This
is fine, so long as everyone knows their place. I've worked in another
environment with a commercial RCS-based tool (in other words,
lock-modify-unlock). This worked because it couldn't fail. Take your pick.
As I said, I'm talking about real-world commercial development; web
development of free software could be seen as being different.
RT
|
 |
Browse
