[plt-dev] VCS poll

Eli Barzilay <eli at barzilay.org> writes:

> * On a scale of -10 to +10, how much do you want to switch to a
>   distributed system?  (-10 => I really want to stay with svn;
>   +10 => I really want to switch; 0 => don't care.)

+10

> * What are your preferences for a distributed system?  

Mercurial
git

> * On a scale of 0 to 10, how concrete+technical are your reasons for
>   having the above preference?

10

I have extensive experience with Mercurial (I switched Scheme 48 from
Subversion to Mercurial, and XEmacs from CVS to Mercurial), which is
well-designed and very well-documented.  (Bryan O'Sullivan did the
book.)  It's also as close in usage to Subversion as a DVCS gets.  It's
used by many big projects (Java, Xen), and very stable.  Mercurial is
nicely stratified: The beginner starts with the basic commands;
higher-level kung fu (like the wonderful Mercurial Queues) has to be
explicitly activated.

git, from what I know, has a model equivalent to that of Mercurial.
I've had to work with it because some projects I contribute to use it.
Every time I do, I find it very painful to figure out what the command
is to do what I want to do.  I've destroyed work on a number of
occasions.  (Google for "svn revert git" and find out how it's done.)  I
think the main problem is that the design of git isn't clearly explained
anywhere - AFAICS, the documentation sucks compared to Mercurial.  (At
the last WG2.8 session, we had an informal session on DVCS - I remember
Norman Ramsey had the exact issues I had.  I believe Olin did, as well.)
For me, the best source of information on git is, perversely, the
translation chart from Mercurial.

-- 
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla