Whether you’re moving up from CVS or getting serious about a revision control system, here’s a powerful tool that will keep records of changes and keep your projects under control.
Revision Control with Arch: Introduction to Arch
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
I’m sorry, but looking at those command line commands in the linked article just makes me glad I use Star Team. Once you’ve used powerful database style software configuration management it is hard to even consider going back to the old days of command line cvs/rcs/sccs.
Sean
As far as I can read from http://bdn.borland.com/starteam/, StarTeam doesn’t provide distributed revision control.
Arch is bad. More precisely, Arch’s command line interface as implemented by TLA is awful. That doesn’t make Arch worthless. Name another distributed revision control system with advanced merging.
I believe this blog entry is relevant here: http://mako.yukidoke.org/copyrighteous/freesoftware/20041017-00.htm…
Quote: “The more you use it, the more you hate it; unfortunately, the more you use it, the more you hate all the alternatives more.”
What if the Linux kernel were saved to distributed version control? Downloading the current version could be nearly impossible. There would be thousands of more or less available hosts. Probably some would be gone due harddisk failure… Somehow distributed version control doesn’t sound very robust.
A Borland product? I certainly loved those, in the days of MS-DOS, but I really doubt it can match the features of more modern tools like BitKeeper or Perforce…
Has anybody used Monotone? It’s supposedly better than arch:
http://www.venge.net/monotone/
Sorry? But the Linux Kernel *is* developed with distributed version control system. It’s called BitKeeper. Do some research yourself please.
BitKeeper is considered to be non-free. (This is complex.) Arch is certainly a free software.
As a side note, FreeBSD is developed using Perforce, which is (like BitKeeper) distributed, and non-free.
Two very big open source projects are using distributed version control system, even if it is non-free. Don’t you think there is a valid reason for this? Or do you believe all Linux/FreeBSD developers are nut?
Are there a lot of people who actually *need* a distributed version control system, and the associated extra complexity? It seems kind of a niche market to me..
Are there a lot of people who actually *need* a version control system like CVS, and its associated extra complexity? Shared folder does the job. It seems kind of a niche market to me.
Well, I am kidding, of course. Learn Arch, and you will think you were kidding.
people use to say that darcs is like arch but slightly more focused on simplicity. I never used it by miself, but maybe worth trying
Well, distributed repositories are certainly not the thing for small projects, and *larger* projects usually *have* a sufficiently powerful central server…
Besides, from looking at the article, it feels like Arch suffers from the Linux UI illness: Needlessly cryptic and complex in the wrong places.
I don’t doubt it’s a powerful system, but the UI sucks. And since it’s the UI you’re working with day-in, day-out, that’s A Bad Thing ™.
But I’m sure someone will write a cozy little wrapper around it (taking away features and adding yet another package to install), and people will claim all is well… :-
> Well, distributed repositories are certainly not the thing for small projects
IMO, they are especially well suited for small projects. The entrance barrier for writing patches becomes low, and contributions mostly help small projects especially.
> *larger* projects usually *have* a sufficiently powerful central server…
And also enough contributors that patches need to be reviewed, and that checkin permissions can’t be handed out to everyone who asks.
> Besides, from looking at the article, it feels like Arch suffers from the Linux UI illness: Needlessly cryptic and complex in the wrong places.
Yeah, tla is way too difficult to deal with. darcs is so much simpler…
None of the preceding posts mention Subversion.
While I haven’t tried it against anything of scale, it seems well fleshed-out, with a (mostly) complete emacs wrapper…
svk is also distributed, and has pretty great star-merging capabilities and other goodies.
It is built on (and interoperates with) subversion. It’s not quite 1.0 yet, and the documentation is kind of sparse, but it’s working really well for me. It’s been a lot easier to get my head around for me than arch/tla.
http://svk.elixus.org
> None of the preceding posts mention Subversion.
No wonder considering the news article is about arch.
Disclaimer: I have never actually used arch, only read the documentation. It’s useles for me until I can use it from WinDos too. That said:
One thing that stands out with arch against the opposition is that it doesn’t seem to be anything “special” at all. It’s an obvious result of putting “diff” and “patch” to good use. It seems so natural that it just *has* to be the global optimum for version control, while the other solutions are just beating around the bush.
It just seems to need lots of polish – a windows client, a better command line interface, and a GUI. After that, it’s set to kill off the competition by being the obvious way to go. Kinda like the wheel is obviously the shape that rolls the best :-).
I heard arch is tied to POSIX, and that a windows client would be very difficult. Has anyone been able to use arch in a windows environment ?
I would have to agree here. StarTeam is a very nice tool and I enjoyed using it. It has some features I have yet to see in other source control tools.
http://abridgegame.org/darcs/
<P>
It’s another distributed OSS revision control system, but unlike arch, it doesn’t suffer from “GNU user interface”. Further, the author has thought carefully about how to implement it, to the extend of providing mathematical proofs, and hasn’t thought it a wonderful idea to write server-software in C and leave everybody to deal with the endless security problems and bugs.
<P>
The only problem with darcs is that it is somewhat slower than arch, but it is improving in that department in leaps and bounds thanks to algorhythmic improvements.<P>
Arch seems to be formed in opposition to subversion, but it is a bit GNUish – ie not all that technically interesting and with an awful interface too and some very dodgy decisions regarding design and implementation. If you really want a distributed system, I’d go with darcs, and if a traditional system is required, svn is the way to go.
>> Well, distributed repositories are certainly not the thing for small projects
> IMO, they are especially well suited for small projects. The entrance barrier
> for writing patches becomes low, and contributions mostly help small
> projects especially.
>> *larger* projects usually *have* a sufficiently powerful central server…
> And also enough contributors that patches need to be reviewed, and that
> checkin permissions can’t be handed out to everyone who asks.
Isn’t this only interesting if you are working on open-source projects with the bazaar development model?
Main development is still done on CVS. Preforce is only used in a supplemental form while doing experimental changes.
[Feel like feeding a troll every once in a while.]
Ummm, you’re a bit clueless, aren’t you? I would ask if you’ve used StarTeam, but it’s obvious you haven’t. I would ask if you’ve used any Borland product since DOS, but it is obvious that you haven’t. In fact, you’ve contributed nothing to the conversation except for revealing your ignorance. I may be wasting bandwidth, but at least I am correcting prejudiced, uninformed drivel.
[Thank you; I feel much better now.]
Posted on Monday, November 01, 2004 by Nick Moffitt
…
hmm apparently we were not supposed to see this for another 11 days.