On 31.03.2009, at 13:17, Göran Krampe wrote:

Hi!

Andreas Raab wrote:

...

Folks - Just as a side note to anyone who is interested in distributed version control systems, I thought it's interesting to see that they get more and more widely adopted. This just popped up on Python-dev: http://mail.python.org/pipermail/python-dev/2009-March/087931.html I'm wondering if this has any impact on our thinking with respect to Monticello, change sets etc?

I am interested in these beasts, and Keith also has quite a bit of experience. We use Hg in Gjallar - although very little. I have also used bzr a bit, no git yet - but it is just a matter of time since Github has really brought git "to the masses". Those are the three "top contenders" although Darcs has a special place in my heart, not for it's magical smarts (well, its cherry picking is way cool of course) but mainly for its tremendous easy of use.

As you may know OpenSolaris, OpenJDK and Mozilla all use Hg (IIRC) and Ubuntu pushes Bzr. These three are all very good, battle tested and fast. Bzr and Hg being written in Python, and git in C.

The Quilt/patch queues in Hg etc is an example of inspiration for Deltas, when Deltas are fully operational such queues can easily be made using them.

IMHO there is a lot to be learned from these tools. Git and Darcs for example represent two different "philosophies" - git following the "don't be too smart, better to be slightly dumb but predictable"- idea and Darcs following the "very, very smart doing magical stuff"- idea. :)

With Deltas I was also inspired by Linus/git - the idea to be able to do relatively nice merging without using history like MC does - thus, "not too smart, but predictable".

I've been working with git for a while now, first in Sugar and lately at VPRI too. I like it very much (compared to svn at least, no experience with hg or bzr), and the best thing is that it is really really simple inside. Highly recommended reading:

http://www.newartisans.com/2008/04/git-from-the-bottom-up.html

- Bert -

"Göran Krampe" goran@krampe.se writes:

Hi!

Andreas Raab wrote:

...

Folks -

Just as a side note to anyone who is interested in distributed version control systems, I thought it's interesting to see that they get more and more widely adopted. This just popped up on Python-dev:

http://mail.python.org/pipermail/python-dev/2009-March/087931.html

I'm wondering if this has any impact on our thinking with respect to Monticello, change sets etc?

I am interested in these beasts, and Keith also has quite a bit of experience. We use Hg in Gjallar - although very little. I have also used bzr a bit, no git yet - but it is just a matter of time since Github has really brought git "to the masses". Those are the three "top contenders" although Darcs has a special place in my heart, not for it's magical smarts (well, its cherry picking is way cool of course) but mainly for its tremendous easy of use.

What Göran said.

I've chosen to recommend Mercurial for my work purposes because darcs likes ssh servers on those machines that will _serve_ repositories, and I've yet to find an easy way to run said servers on Windows machines.

That cherry-picking business of darcs', though: man, that just rocks. When you say "darcs record" it walks you through the hunks in a file, and you can just say "yes, I'll record this hunk, no, not that hunk". (I'd dearly love to see something like this in MC!)

Serving Mercurial repositories is more effort than darcs, but it's not rocket science. I had an apache server anyway, so it was really just bunging hgwebdir.cgi in the right place and writing the config file.

frank

2009/4/7 Göran Krampe goran@krampe.se:

Note that Mercurial has this "chunk wise" recording now too, by default I think. But that is *not* what I meant with cherry picking - but I presume you know that, I meant the concept that you can pull patches "out of order" from another repo. AFAIK Darcs is alone in this, although the others have "similar inferior mechanisms" like transplant, rebase etc.

That's without counting with MC2 IIRC.

On Tue, Mar 31, 2009 at 4:44 PM, Frank Shearar frank.shearar@angband.za.org wrote:

That cherry-picking business of darcs', though: man, that just rocks. When you say "darcs record" it walks you through the hunks in a file, and you can just say "yes, I'll record this hunk, no, not that hunk". (I'd dearly love to see something like this in MC!)

I thought MC2 could work from change sets? If so, you have this now and with an even handier interface. Just arrange the change sets how you want them with the change sorter, then commit it with MC2. Am I correct?

Well the change set mechanism lets you create new change sets and move individual changes between them however you like. Then I would assume MC2 has some way to select which change set to record. I don't know how it works in practice but it could be something even nicer than what Darcs has from a UI point of view.

On Wed, Apr 8, 2009 at 12:19 PM, Frank Shearar frank.shearar@angband.za.org wrote:

Jason Johnson wrote:

...

On Tue, Mar 31, 2009 at 4:44 PM, Frank Shearar frank.shearar@angband.za.org wrote:

...

That cherry-picking business of darcs', though: man, that just rocks. When you say "darcs record" it walks you through the hunks in a file, and you can just say "yes, I'll record this hunk, no, not that hunk". (I'd dearly love to see something like this in MC!)

I thought MC2 could work from change sets?  If so, you have this now and with an even handier interface.  Just arrange the change sets how you want them with the change sorter, then commit it with MC2.  Am I correct?

I've not used MC2, so if you can selectively commit parts of what you've changed, I'd be very happy. Or, to make me just as happy, point out the MC1 feature I'd missed.

frank

PS I'm _trying_ to get Thunderbird to send this using text/plain; apologies if I fail.

Jason Johnson wrote:

On Tue, Mar 31, 2009 at 4:44 PM, Frank Shearar frank.shearar@angband.za.org wrote:

...

That cherry-picking business of darcs', though: man, that just

rocks. When

...

you say "darcs record" it walks you through the hunks in a file, and

you can

...

just say "yes, I'll record this hunk, no, not that hunk". (I'd

dearly love

...

to see something like this in MC!)

I thought MC2 could work from change sets? If so, you have this now and with an even handier interface. Just arrange the change sets how you want them with the change sorter, then commit it with MC2. Am I correct?

Sort of. Yes, you could do that.

But if what you want is cherry picking, MC2 supports that with packages as well. When doing a merge you always have the option to choose which changes to apply. A normal, non-cherry-picking merge is just the degenerate case where you choose to apply all the changes.

Colin