CVS "just works?" (was RE: Bazaar book from Packt)

Jelmer Vernooij jelmer at samba.org
Sat Sep 7 10:05:37 UTC 2013


On Sat, Sep 07, 2013 at 02:43:48AM -0700, Mark Grandi wrote:
> I don't see how colocated branches help with maintaining a large and
> active product, or were those two separate things

Colocated branches make it a lot easier to work with just one (or only
a couple) of working trees and a lot of branches. Apart from the
source files, this also means that you don't need to rebuild all
object files for each branch, but just those that have changed.

Even without colocated branches, there are ways in which you can share
a working tree with multiple branches but this is a fair bit harder
and requires more setup.

Cheers,

Jelmer

> From: Stephen J. Turnbull
> Sent: 9/7/2013 1:23 AM
> To: JP Vossen
> Cc: bazaar at lists.canonical.com
> Subject: CVS "just works?" (was RE: Bazaar book from Packt)
> JP Vossen writes:
> 
>  > I totally agree, which is why I find everyone else's love of git so
>  > confusing.
> 
> Not everybody loves git.  The thing about git is that it opens up to
> all users the possibilities of DAG manipulation that Bazaar offers
> only to its core developers and a few enthusiasts.  That comes at a
> cost of git's surface complexity and annoyance to people who want it
> to "just work" (or better yet, "just go away" :-).  That is probably
> the great majority of all programmers.
> 
> But you shouldn't be surprised that a large fraction of the open
> source community loves git.  One good reason is that git *does* open
> up its core functionality to the user: we like our tools sharp and we
> hone them ourselves.  Sure, Bazaar is open source, but I have neither
> the skills nor the equipment to forge a replacement blade starting
> from iron ore.  With git, it's more at the level of whetstone and
> sharpening oil, and maybe a bench vise.
> 
> Another reason is that git is the best tool available for handling the
> kind of massive, distributed branching that occurs in a hyperactive
> public project (and I'm not just talking about the Linux kernel).
> Colocated branches really help.  Mercurial has colocated branches, but
> the implementation sucks by design.  Bazaar's implementation isn't
> supported, and its UI ain't "there" yet.
> 
>  > I do not want to have to know as much about the guts of the
>  > tool as git requires in order to not shoot myself in the foot,
> 
> What need to know "guts"?[2]  You "just" need to be disciplined about
> what commands and options you allow yourself to use.  But that's a big
> defect when git sports a plethora of non-orthogonal commands and
> options, and often you *do* need to use options, default behavior
> isn't quite right.  That's very unattractive to "I want it to just
> work" users.
> 
>  > and I *certainly* don't want to have to try to explain all of it to
>  > other folks who don't understand or want to use a VCS in the first
>  > place!
> 
> Actually, I find those folks to be among the easiest to introduce to
> git.  Eg, my economics students.  They only know five commands: init,
> add, commit, push, and log.  If push came to shove, *they* could get
> along without log.[3] :-)  The hardest part of getting them going with
> git is getting ssh installed and configured on Windows.  I wouldn't
> care what dVCS they use in this application, but if it's git I can use
> gitolite (a tool I use anyway) to manage communication.
> 
>  > [...]
>  > > We still use CVS because it "just works", doesn't get in the way.
>  >
>  > Unless you tag every single commit, or change only a single file at a
>  > time, I have to disagree there.  CVS is fundamentally broken because of
>  > the way it does everything "per file."
> 
> There are a lot of applications where CVS and RCS are fine.  Eg, my
> students would do fine with either; they don't branch at all, and
> often they only have one file under VCS.  The choice of a modern dVCS
> is dictated by the "d": they can work offline and submit for review
> with one simple command the next time they're online.
> 
> 
> Footnotes:
> [1]  With the single exception of git-gc.
> 
> [2]  Not that the guts are difficult.  They are just a slight
> generalization of singly-linked lists as used in the "guts" of Lisp.
> 
> [3]  And they don't use "diff" because they all use Word to edit their
> documents. :-P
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <https://lists.ubuntu.com/archives/bazaar/attachments/20130907/dbb5a262/attachment.pgp>


More information about the bazaar mailing list