[Yuval Levy <sf05@sf...>]

On January 28, 2011 12:09:47 am D M German wrote:
> here are some reasons I like git over mercurial (please keep in mind i
> am not a hg expert, so I might be confusing/missing something).

I really don't care which one it is between git and hg.  DVCs are more or less 
interchangeable.  I know of at least one dev on the Hugin team that used git 
for his personal repo while Hugin was still SVN; and interfacing between hg 
and git is even easier.

I would prefer not to have to install all of them; but it is a weak preference 
and I can live with any of the envisioned tools.

 
>   - I like that there is a single end-node in git (the head) but I don't
>     like that hg allows multiple ones (heads) and one of them is the
>     tip (this might be an advantage of hg, but I am not
>     sure because I haven't used it).

A hook can be configured to prevent multiple heads [0]

 
>   - I like that everything in git happens in one single directory, where
>     you move across branches with one command, and you can "stash"
>     changes. In mercurial, every branch is a subdirectory of the
>     repository.

I am not sure I follow you here.  With Hg, everything happens in one single 
directory as well.  Within that directory, I can bring up any branch with

   hg up -C <YOUR_BRANCH>

and I can copy any changeset from any other branch to that branch with

   hg export --git <REVISION_NUMBER> | hg import -

note the --git flag which uses the git format for diff.  I have not tried to 
move between Hg and git, but I would not be surprised if it was very easy and 
you and I could develop the same codebase in the two tools, pushing/pulling 
changes from each/other while you use git and I use Hg.

 
>   - Git has more traceability: I like how committer and author are both
>     recorded in git. In hg, committer info is lost.

I can't really comment on this, I don't know Git well enough.  Traceability is 
sure a weakness in hg, I can "spoof" the committer and author information all 
along the way.  Then what is the point of differentiating committer and author 
if that difference is spoofable?

 
> - Operations (important)

my take on that is that functionality permeates across DVCS:  enough users are 
smart and capable coders and will copy from each other:

 
>     - git-grep (which can search history),

[1]

 
>     - git-bisect (binary search of the change that introduced a bug),

   hg bisect


>     - git-diff, which pinpoints the function where the change is
>       happening (I don't think hg does this).

If it is "just" the diff format that is different, hg diff --git should do it. 
[2]
 
 
> Other than because "hg is used by hugin", can anybody comment on why hg
> rather than git?

The reasons why I like Mercurial:
* documentation
* it is scripted and very easy to extend/customize/modify
* if you have access to some basic webspace, you can do without central repo.


> Now, libpano does not have many developers, so what I really want/need
> is easier management of experimental (mostly personal) branches.

In that case I would go back to the original argument (I think by Tom): why 
change for change sake?  Gerry Patterson set up his own git repo to interface 
with Hugin's SVN back in the days when Hugin was still on SVN.

Yuv


[0] 
http://mercurial.selenic.com/wiki/TipsAndTricks#Prevent_a_push_that_would_create_multiple_heads
[1] http://blog.axant.it/archives/283
[2] http://mercurial.selenic.com/wiki/GitExtendedDiffFormat