RCS Comparison table and ease of use

John Arbash Meinel john at arbash-meinel.com
Wed Feb 7 18:52:39 GMT 2007


Nicholas Allen wrote:
> Hi,
> 
> I'm curious why bazaar has a "No" for ease of use in the comparison
> table. It is one of the easiest to use RCS systems I have ever used.
> Personally I think this should be a big "Yes" ;-)
> 
> Nick
> 
> 

Well, you would have to ask SoloTurn, who made some comment-less edits
in this range
http://bazaar-vcs.org/RcsComparisons?action=diff&rev2=82&rev1=76

There is only one reference, really, for bzr, and it is a Bazaar versus
darcs comparison.

It is a little hard to read
http://www.kdedevelopers.org/node/2024

Because it seems to be confused about Baz 1.x (the Arch project,
formerly known as Bazaar) versus Bazaar (bzr, fka Bazaar-NG)

As near as I can tell, that kde developer preferred interactive commands
(which we feel don't really fit the unix philosophy. ATM we have only 1
interaction that I'm aware of, and that is uncommit)


I have personal feelings about Bazaar's ease-of-use, but as a main
member of the community I feel it is shameless self promotion to edit
that line.

If you as a user and not a developer feel differently. Please update our
Wiki. Even better if you can have a blog post that describes how easy it
has been for you to use Bazaar.

I will say, that versus darcs, we are probably a little bit more
difficult to use.

Though I should point everyone to Bryan O'Sullivan (mercurial developer):
http://www.serpentine.com/blog/2007/01/11/why-i-dont-like-darcs/


There are also specific design choices. Like interactive commits. We
explicitly don't do them, because we feel it is better to have well
tested code that you are committing, not just random chunks.

We have evolved 'bzr shelve' as a way to split up your current tree into
chunks before you commit, though really 'shelve' should be in core
before we really point to it too much.


I actually think that is a decent post as to why there isn't an absolute
statement for why "Ease-of-use" is really too subjective to be put into
a Yes/No statement. Especially if you read all the follow up comments.

There have also been git vs bzr discussions of whether Branch + revno is
actually easier than git's sha-hash numbering. There are tradeoffs to
both. Obviously we personally believe the trade-offs favor branch + revno.


There is also a question as to whether a bug like this:
http://darcs.net/DarcsWiki/FrequentlyAskedQuestions#head-76fb029ff6e9c20468eacf3ff00d791e2cf03ecb

Means that darcs is/isn't easy to use. (If I was using a tool that made
everything easy, but every so often it blows up and is completely
unusable, is that easy to use? Maybe if the blow-up is rare, and easy to
work around. Maybe not.)


In some ways I would rather remove that entry, because it is a very
subjective evaluation. Instead, I would rather just have a collection of
links at the end of the document where people have compared different
RCSs and discussed their feelings. People can read them, and get their
own impressions.


(Another ease-of-use with darcs.  How do I get my tree back to an exact
state in the past if I didn't tag it? Is that even possible? I have the
feeling darcs developers don't feel like this is an important operation.
And probably quite a few people agree.)

John
=:->



More information about the bazaar mailing list