Thank you, Bazaar designers, for first-class rename operation | Hacker News Reader

Thank you, Bazaar designers, for first-class rename operation

old.nabble.com

10 pointsprog12 comments

I'm also using bazaar, after moving from SVN. I selected it over git or hg because 1) it is very adjustable regarding using it either in a centralized or distributed setup, working over ssh, even without any server-side support and 2) its command are very easy to learn and the branch and version numbering is best of the 3, IMO. Git was very foreign for me, coming from the SVN.

bzr rename, true directory support, and made working with Subversion more sane. All the reasons I switched to bazaar from Mercurial.

Now if I could get the trunksapp.com guys to support Bazaar, I'd be golden.

Anyone have a link to toss in, or a comment, on a few issues ? Namely:

- Binary diffs (svn is best was my impression)

- Multi-platform (svn and hg do well ?)

- Hg seems to get the most non-git DVCS uptake. Are the bzr technical merits enough to justify the push by Canonical/Ubuntu ?

P.S. just took a look at http://bazaar.canonical.com/en/ ; I had not checked on bzr for a couple of years, and it looks impressive now, with plenty of multi-platform support and nice-looking GUI.

> - Binary diffs (svn is best was my impression)

I am not sure how svn handles binaries.

I believe bzr uses binary diff ( http://old.nabble.com/Re%3A-SVN-and-graphics-binary-files-p2... ) IIRC there is also a file locking plugin.

> - Multi-platform (svn and hg do well ?)

As you already mention (as PS), multi-platform support is good. I primarily use it on Windows and Linux. Haven't tried OSX but there is an installer.

> - Hg seems to get the most non-git DVCS uptake. Are the bzr technical merits enough to justify the push by Canonical/Ubuntu ?

I haven't use hg much so I can't comment on that but bzr does have better storage efficiency as mentioned at http://doc.bazaar.canonical.com/migration/en/why-switch-to-b... . Its closer to git in repo size. rename and directory versioning is first class (no need for .ignore files). Performance is comparable to hg. One reason I use bzr is its support for wide range of protocols including sftp, ftp. There are some good plugins like bzr-upload[1] that allow working with remote hosts. bzr-explorer[2] is also nice.

Also, launchpad integration is really good.

[1] https://launchpad.net/bzr-upload

[2] http://doc.bazaar.canonical.com/explorer/en/

Push? Bzr is sponsored by Canoncial.

To me, it even seems mostly made by Canonical (they go into the issue). What I meant is: to they have a technical point, or is it mostly NIH/stubbornness ?

(The desk I'm using right now has an Ubuntu machine, and a Mac; there are quite a few things in which Ubuntu is more pleasant to use than the Mac. In a few years, we may be saying that Mark Shuttleworth is the "Steve Jobs" of Linux, in a good way of course ;-)

Bzr is a huge and slow mess, I can't imagine why anyone would switch to it from git or hg.

I've had no problems with speed. How big of a repository are you working on? The largest I work on is about 70MB of source (code, html, images, documentation, etc.). I use a shared repository and branches take maybe a second or 2 to create.

Thats very informative. Thanks.

Attempt to clone Emacs' repository (http://savannah.gnu.org/bzr/?group=emacs) sometime and you'll understand he's being very informative.

Or I could stick to the launchpad mirror[1] which is extremely fast. savannah uses dumb http for bzr[2] and hence it is slow. I have been using the launchpad mirror for some time now.

[1] https://launchpad.net/emacs

[2] http://article.gmane.org/gmane.emacs.devel/121222

Is it still slow in v2.0+ ? They quote 1.0 sec commits on the Firefox 3.5 repo test (vs. 1.1 for hg, 0.35 for git). ( cf. http://doc.bazaar.canonical.com/migration/en/why-switch-to-b... )

The http://whygitisbetterthanx.com/ site, sportingly linked to from the top of the page above, also has some comparative numbers (must click headers to show), but I don't see version numbers of the programs tested ...

Reply on news.ycombinator.com