[Arx-users] Further thoughts on ArX and simplicity

[Kevin Smith]

Greetings all,

For the last couple months, I have been immersed in the git community, 
and then the mercurial community. In case you're not following those, 
git is Linus Torvald's super-minimalistic BitKeeper replacement, and 
mercurial is another new distributed SCM.

I have realized that git really doesn't meet my needs, for a variety of 
reasons. The main appeal of mercurial for me is that is presents a 
simpler UI than ArX. Mercurial is still a contender, but I am now 
feeling drawn back to ArX for a couple reasons.

This email is to explore possibilities for further simplifying ArX.

The single biggest usability flaw in ArX right now has to be the 
confusion around URL's, archive names, and branches. Plus the fact that 
you have to specify one or more of those far more often than you do in 
mercurial.

Walter:

You proposed this new naming policy:

  archive#branch.subbranch,revision

This sounded like a good idea at the time, and I still think it would be 
a huge improvement. In my mind, it transforms one of the worst features 
of ArX (confusing naming) into an advantage, because other systems would 
tend to say something like archive -b branch.subbranch -r revision.

You also talked about assigning a GUID to each archive, as a tool that 
would allow you to simplify the archive naming and relationships in 
other ways. It's still not quite clear to me why a lightweight design 
wouldn't work instead. Something like:

- Archives are identified by a URL
- The user can optionally set up aliases for archives
- "Lightweight branches" should generally be tied to aliases, to allow 
following the upstream source if it moves

With that, nobody has to worry about (be confused by) GUIDs. Everything 
is simple and transparent. Archives don't have names, and and I think we 
could remove the special case of "mirrored" archives. Mirrors would just 
be a plain copy, at a different location. Users can set up aliases if 
they wish, or they can use raw URL's.

Whenever possible, the archive should be inferred based on the current 
working tree you are in. Whenever possible, the branch should default to 
a reasonable value.

Would you accept patches to move in this direction?

Kevin