Jump to content

Talk:Distributed version control

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Distributed vs Decentralized

[edit]

I am a bit confused, why does this kind of thing have label of "distributed". According https://wiki.riteme.site/wiki/Distributed_computing definition, it means that more than single computer is cooperating on single operation. This does not apply to "Distributed version control". I would personally call it "Decentralized version control". Where is origin of "Distributed" come from? PS: As not native speaker I am not sure about etymology of given word. This can perhaps help too.

Can someone add a few sentences about this discrepancy to the article? — Preceding unsigned comment added by 81.92.248.121 (talk) 09:30, 4 July 2021 (UTC)[reply]

What does this mean?

[edit]

"to work on a given project without requiring that they maintain a connection to a common network." This doesn't make any sense. What is a "common network" entailing? How does one work on a "given project" without requiring that they "maintain a connection to a common network." make sense? — Preceding unsigned comment added by 169.139.19.96 (talk) 20:41, 13 March 2014 (UTC)[reply]

Without looking up in the article what exactly you're quoting (but interpolating a fragment from your second question), your sentence fragment looks like there must have been a preceding "A DVCS enables several developers to work on a given project without requiring..." (or some such). Given this, the answers to your questions are:
1) "A common network" is a network that these developers are all connected to. Since this is not required (see the next answer), there does not need to be such a network.
2) The developers work on a given project without requiring that they maintain a connection to a common network by each working on their own local clone of the code repository. This frees them from any requirement to be constantly connected to the network where the code was originally stored, i.e. from maintaining that connection.
(They do of course intermittently need to connect to it: At least once at the beginning, to create their own local copy; and then optionally later to integrate their changes "upstream" or to fetch updates uploaded by others. But this is not mandatory, and anyway doesn't require a maintained -- i.e, always-on -- network connection.) --CRConrad (talk) 11:08, 10 June 2022 (UTC)[reply]
  1. NULL
2600:1700:F6E0:4750:71F4:CDD0:83F9:FDD (talk) 22:11, 17 May 2023 (UTC)[reply]

Cleanup

[edit]

I just tagged this article for cleanup, but sourcing is another issue. Please don't be offended if you have been working on this article. I have put the tag up for the following reasons:

  1. The article is rather 'listy'.
  2. the article is mainly unsourced
  3. The section headings don't follow the Manual of style.
  4. There are many external links in the article, that should be replaced with wikilinks, or not be linked at all.
  5. The statement "fairly recent" in first paragraph is not a statement that will age well. Give an approximate year at least.
  6. The introductory part after "other differences are as follows" is confusing organizational issues (peer approval, and all that Linus styled organization) with the actual DVCS differences with centralized VC. After all, you can organize centralized repository access in same manner, even distributed repositories.

If someone could provide some sourced from which this article is built, I'll be happy to help out improving it. Martijn Hoekstra 19:12, 2 November 2007 (UTC)[reply]

Merger proposal

[edit]

I propose Distributed revision control & revision control#Distributed revision control should be merged in one way or another.--Grimboy (talk) 17:27, 1 January 2008 (UTC)[reply]

Comment: If the merge happens, I think it should be from this article to Revision Control, not the other way around. RossPatterson (talk) 20:08, 1 January 2008 (UTC)[reply]
Comment: I agree with RossPatterson. Whether or not distributed revision control needs a separate article, the main article needs to cover distributed revision control. Both topics gain from close comparison. If merging, merge from Distributed revision control to Revision control#Distributed revision control. --Michael Allan (talk) 21:45, 1 January 2008 (UTC)[reply]
Comment: Merging is a bad idea. This part just tells readers DVC exists and a little about how it differs. This is 1005 legit. and nec. for a rounded view of the topic of VC. Very fine details of course belong in DVC's own entry. Removing this could potentially leave the reader ignorant of DVC. —Preceding unsigned comment added by 24.12.136.199 (talk) 02:23, 29 December 2009 (UTC)[reply]


Comment: I disagree with the merger. Maybe its just my preference, but I feel that good articles need some degree of overlap with their sub-topics, with each of the replicated parts displaying different levels of detail. The lower level of detail in the overview should introduce and summarize the more detailed sub-topic referenced by in the hyper-link. This reduces the tendency for reading hyper-linked texts to become disjointed, whereby the reader becomes lost in an endless regression of definition, all at full detail. I think this section should remain as an introduction to the linked and more detailed sub-topic. --Daniel Collicott 09:55, 7 February 2009 —Preceding unsigned comment added by 78.105.184.242 (talk)
(1) Errm, I don't get this. Someone suggests a merge; a couple of people agree; after a year nobody has disagreed, and yet the merge has not taken place?
(2) I think a merge would be a good idea, but if not then this article needs completely rewriting, as it is appallingly badly written. For a start, nowhere does it tell us what "Distributed revision control" is. JamesBWatson (talk) 19:30, 22 April 2009 (UTC)[reply]


I agree that the articles should be merged. It is a huge problem when a single topic is broken down into pieces placed on separate pages. The purpose of a encyclopedic presentation of information is to find that information in one place. A longer page is not a problem, but separate pages are. Daniel's objection may well be solved by the proper use of XOXO within wikis like this. But that is a separate issue. In the mean time, I believe that the use of clear sub-headings will solve the problem. - KitchM (talk) 18:19, 20 October 2009 (UTC)[reply]

Diagram

[edit]

A diagram of some kind illustrating the differences between centralized and distributed VC would help a lot —Preceding unsigned comment added by 83.89.0.118 (talk) 15:28, 23 June 2008 (UTC) Exactly! --62.153.161.241 (talk) 07:37, 18 January 2010 (UTC)[reply]

History

[edit]

The article says that "The first DVCS is Reliable Software's Code Co-op (1997)". However, I've just run across an academic paper describing one earlier than that: 'O'Donovan, B.; Grimson, J.B., "A distributed version control system for wide area networks," Software Engineering Journal , vol.5, no.5, pp.255-262, Sep 1990'. It was implemented over UUCP (!) although to be fair, they say they wrapped UUCP calls in an abstraction layer so it could be replaced. Should this be added to the history section? —Preceding unsigned comment added by Ajb (talkcontribs) 13:39, 28 July 2008 (UTC)[reply]

Sure, that's really intersting. RossPatterson (talk) 13:01, 15 August 2008 (UTC)[reply]


I started using Sun Microsystems TeamWare around 1990 or 1991 while working for Cray Research on their CS6400 computer which ran Solaris 2. Our team ported Solaris 2 to the CS6400 and used TeamWare because that is what Sun used for the OS Networking source base. TeamWare was a full DVCS even back then. I'm not sure how long Sun had been using it before then. Does anyone here know how to contact Larry McVoy (who designed TeamWare and BitKeeper)? He would probably have much more detailed knowledge of the history of DVCS. In any case, it was a long time before 1997.

Billdav (talk) 19:30, 30 March 2009 (UTC)[reply]


First generation open-source DVCSes include Arch and Monotone. The second generation was prompted by the arrival of Darcs,

It is not clear, in which sense the DVCSes are names first and second generation. There are no other mentions on the page. Also, from brief search it seems to me that Arch origined in 2001, Monotone in 2003, Darcs in 2002.

Dendik (talk) 20:28, 28 September 2009 (UTC)[reply]

locking?

[edit]

hi guys,

might locking be a feature people used to svn might miss in dvcs. i see no way of having lock-modify-commit in dvcs but arts people only want it this way. —Preceding unsigned comment added by 195.143.104.254 (talk) 14:08, 19 January 2010 (UTC)[reply]

Disadvantages are just FUD

[edit]

The "disadvantages" section contains a couple of myths about DVCS that aren't actually true. They aren't harder to use, on the contrary their key selling point is that they make the hard things (branching and merging) very easy. Also, cloning a git or hg repository isn't much slower in practice than a Subversion checkout. Maybe by a factor of two or so at most, but no more, and besides, it's only the kind of thing you do occasionally so it's pretty much a non-issue. 194.60.38.198 (talk) 08:36, 17 June 2010 (UTC)[reply]

Not everyone in the world has fast internet and "factor of two at most" is huge when it is a matter of 1 hour vs. 100 hours for a huge project. Your FUD allegations are just arrogant. — Paul Pogonyshev (talk) 11:01, 12 July 2010 (UTC)[reply]
One hour versus 100 is a factor of 100, not a factor of two, and you'd be looking at seriously huge projects to see anything near that kind of difference. Incidentally, the other evening I compared a Subversion checkout of the Django project with a clone of the Git mirror on github. Despite the fact that it's over 13,000 revisions, I found that git cloned the entire history in half the time taken by a Subversion checkout. 194.60.38.198 (talk) 11:56, 16 July 2010 (UTC)[reply]
He means.. if your project takes 1 hr to checkout, then taking 2 hrs is probably annoying but can be lived with. If it already takes 100hrs then doubling that adds an extra four days to your schedule. Exxaggeration maybe, but valid none the less, operation speed matters a lot to some people. If you use continuous integration this can become very important because it affects your peak integration speed. EasyTarget (talk) 13:51, 28 October 2010 (UTC)[reply]


firstly, thanks to EasyTarget for explaining to the user 194.60.38.198 what Paul Pogonyshev meant. Hehe. I also disagree to a merge. the revision control article really meets Wikipedia 'Standards'. i.e. its written with references (or source) thus, its more believable/factual. the DVC article is written without references and written in a 'rebellious' style. Ofcourse DVCS are 100% legit.But this article is not.! Also, the info. given in the main article(revision control...) about Distributed revision control systems is enough. to user JamesBWatson- i think these are the reasons why "the merge has not taken place"

                  -------Rampuse 59.96.88.156 (talk) 17:05, 27 January 2011 (UTC)[reply]
Well the current wording "...disadvantage of DVCS, one could note..." is awful: a real 'wonks grudging acceptance that this favourite toy might have a blemish'.
No system is perfect, things that make DVCS systems really great for freewheeling distributed agile/scrum developments can make it really hard work for other models.
EasyTarget (talk) 13:52, 28 October 2010 (UTC)[reply]
I think that a general speed difference is difficult to hold. Git, for example, is a very, very fast system (git commands on a native Linux system with modest hardware take typically times of a hundreth of a second), which is usually much faster than non-distributed version control systems such as Microsoft Team Server. Also, a speed difference in the initial pull of a repository is in almost any practical case not relevant at all, because it will either be much less than a second or the project is exceptional large and the time to understand a project and get it running the first time will be much much larger than the time required for the transmission.
The other critic is that binary files cannot be locked. However, locking leads very often to problems and solves very few problems. It is possible to construct cases in which locking will be needed, such as an image processing company with many people working in parallel on the same binary files, however in software development, this just does not apply. --84.135.67.155 (talk) 22:06, 27 May 2011 (UTC)[reply]

"Every working copy is effectively a branch."?

[edit]

In my wording would be "Every working copy is effectively a fork." As a copy does not necessarily mean commits and commits are needed to talk of a branch. Branching happens whenever one revision has more than one successor. Changing that in the article ... —Preceding unsigned comment added by 188.174.24.74 (talk) 11:22, 10 March 2011 (UTC)[reply]

pull request

[edit]

Pull request redirects here but there's nothing in the article defining the term. Looks like a TODO. 2.101.108.194 (talk) 11:59, 16 October 2012 (UTC) , pbhj[reply]

Yep - I second this - same thing just happened to me.Iamsorandom (talk) 13:39, 16 April 2013 (UTC)[reply]

Me too - now it's almost 2 years ... JB. --92.195.48.144 (talk) 15:05, 15 April 2014 (UTC)[reply]

It's not really a technical term or VCS feature. It just means you make a patch in your local repository and then ask someone maintaining an upstream repository to pull your change. It's like the "edit request" you made here on Wikipedia by posting to the talk page. Github has an actual pull request button, but again, that just posts some alerts for people to look at. Added: I guess LKML has a bit of formal process around pull requests, but I'm not a participant and am not sure how it works. 70.36.142.114 (talk) 16:47, 13 May 2014 (UTC)[reply]

The current explanation remains unclear. If "pull" means "import the code and incorporate the suggested changes", please explicitly say this. I would edit the section to say this myself but I am not sure that this is what is meant. -- Newagelink (talk) 05:46, 21 May 2017 (UTC)[reply]

Distributed version control vs. Distributed revision control

[edit]

Why name this article "Distributed revision control" instead of "Distributed Version Control"?

It is interesting to see that if you lookup that expression on Google, apart from this article and the Revision control article, the first page of results is mostly pages on DVCS, not DRC or DRCS.

On a more quantitative and objective note, Google Fight place DVC an order of magnitude higher than DRV (litterally: 9,960k against 979k at the time of writing). Nowhere man (talk) 17:10, 18 May 2013 (UTC)[reply]

Expert opinion

[edit]

I'm hoping an expert could polish the prose on this article. Or perhaps suggest a merge with Revision control? Given the separate treatment of this topic, the current structure seems ripe for rethinking. — HipLibrarianship talk 01:48, 28 August 2014 (UTC)[reply]

Experts are not needed for prose polishing or proposing a merge, so I'm removing the tag. RockMagnetist(talk) 02:24, 4 September 2015 (UTC)[reply]
  • Merging this into revision control would be an awful idea. Version control has decades of history behind it, but distributed revision control is much newer (8-9 years?). It's a different approach and wikipedia really needs a better article to explain it. The current one... doesn't. "Experts are not needed" indeed! Like that has worked.
Also why is this called "distributed version control" rather than "distributed revision control"? That's reasonable, although a bit out of date, for version control, but it gives quite the wrong idea when applied to the current generation of distributed (Git-like) systems. Viam Ferream (talk) 15:45, 20 January 2016 (UTC)[reply]
See below: GIT is not the inventor but just one of the BitKeeper SCCS clones with an own history fileformat. Schily (talk) 11:06, 21 January 2016 (UTC)[reply]
That ship has sailed - see the discussion at Talk:Version_control#Requested_move_13_July_2015. RockMagnetist(talk) 20:53, 20 January 2016 (UTC)[reply]
Yeah, one old guy sayinbg "I invented this back in the '80s and punch cards were good enough for my granpappy an' me, so that's how it's staying" What relevance does SCCS still have to anything distributed? Viam Ferream (talk) 09:37, 21 January 2016 (UTC)[reply]

SCCS was the first system that supported distributed version control and SCCS is the model for all other modern systems:

  • Together with NFS in 1985, SCCS was made aware of locks from other NFS clients by Sun, supporting contemporary usage in a group of NFS mounted clients.
  • In Spring 1986, NSE was published by Sun - a frontend to SCCS based on the Translucent Filesystem (together with various SCCS enhancements) it enabled distributed usage.
  • Around 1990, Teamware was published by Sun - a frontend to SCCS based on NFS (together with more enhancements to SCCS) it was e.g. used for the distributed development of CDE by Sun, HP, AT&T, IBM, ...
  • In the late 1990s, Larry McVoy a former Sun employee from the Teamware team reimplemented SCCS under the name BitKeeper and added distributed features based on TCP/IP and mailed commit deltas.
  • In 2005, Mercurial and GIT reimplemented the user interface from BitKeeper and even support the commit delta format from BitKeeper
  • In 2006, the original SCCS sources have been made OpenSource by Sun (after Caldera Linux prevented a similar deal with SCO in 2001). Since then, SCCS is under active development and soon will be available with integrated distributed features.

I hope it helps to check the past to understand the current situation. Schily (talk) 10:55, 21 January 2016 (UTC)[reply]

SCCS is no more distributed than many others, like Subversion. It allows distributed access to a repository - as does Subversion - but what Git does is more than that.
How would you use SCCS to work in a Git-like way, where every repository is its own fork, there is no shared access via filesystem or server, and you can move revisions (changes, not versions) easily between them? You can't. This is what distributed revision control systems do, which the previous generation of version control systems couldn't. This is why these two articles shouldn't be merged.
If SCCS was "hiding its Git under a bushel" all this time, then why isn't it in this article? (News to me, but maybe someone did implement this same stuff years ago). If SCCS is just a server or networked filesystem granting distibuted access to ONE CENTRAL REPOSITOPRY (as I think it is, and as Subversion is) then that's a different thing. Viam Ferream (talk) 14:43, 21 January 2016 (UTC)[reply]
I recommend you to take a breath and to carefully read my text again. Your reply was not helpful at all. Schily (talk) 14:53, 21 January 2016 (UTC)[reply]
Maybe we're just disagreeing on what "distributed" means. I see there as being more to it than just NFS.
Do you think Subversion is "distributed"? Do you think Git (and that generation) are different to this, having gained a new feature that simply wasn't there before?
If you see those two as being different, where do you see SCCS as being? I see it as Subversion-like, not with Git's features.
If SCCS really did "do everything Git does, years earlier", then why did we bother with CVS or Subversion? Why all the hoopla when Mercurial and Monotone appeared? Viam Ferream (talk) 15:07, 21 January 2016 (UTC)[reply]
I would agree that SCCS is "distributed" in the sense that it provides access from across a network.
I accept that the internal file formats are inherited from some earlier product (I have no idea either way - I've never looked).
The "Git generation" though do something in the middle that is novel. They allow multiple repositories to each be their own fork, and for this to be the general way of working. There is no "HEAD" or "trunk" branch (as for Subversion). There is no need for a convoluted and manually-intensive merge process to re-synchronise changed repositories. All repositories are truly peers, no one repository has priority as the "master" (unliek Subversion - Subversion's local copies are always subservient to that central repository). This much is novel. Viam Ferream (talk) 16:41, 21 January 2016 (UTC)[reply]
Please do not missread my text a second time.
BitKeeper SCCS definitely is a distributed version control based on the ideas that have been developed with front-ends to the classical SCCS implementation.
The classical SCCS implementation on the other side is underway to become a distributed version contol soon. If you did carefully read my previous test, you could know this already. Schily (talk) 17:44, 21 January 2016 (UTC)[reply]
Bitkeeper isn't SCCS. Now I would include Bitkeeper with Git and Monotone as one that is distibuted, but that doesn't apply retroactively to any old version of SCCS, as you seem to be claiming! Viam Ferream (talk) 09:33, 22 January 2016 (UTC)[reply]
@Viam Ferream: The "revision" vs "version" debate at Talk:Version_control went through multiple rounds and was eating up a lot of the editors' time. I made the decision as an uninvolved admin (I don't really care which term is used). It was an easy decision because Wikipedia content is based on sources and the sources overwhelmingly favored "version". You need to realize - discussions like the one above will change nothing because they are just narratives, not discussion of sources. RockMagnetist(talk) 18:12, 21 January 2016 (UTC)[reply]
That's a different article though. Is this the popular wikipedia idea again that "all vaguely related terms must use identical words, because consistency is more important that accuracy"?
I would agree that version control is right, because that is the popular term used and has been for years. Distributed systems are distinct. General use of the term here is much more evenly split and there's also a new concept that wasn't there before: the idea that the revison (ie the dynamic change made) becomes the interesting term for the user to consider, not the static version. We should highlight this, in this article at least. Viam Ferream (talk) 09:33, 22 January 2016 (UTC)[reply]
@Viam Ferream: If you don't read my replies, it does not make sense to continue this discussion. I did never claim that the SCCS implementation from Marc J. Rochkind was already a distributed version control, but there have been frontends 30 years ago that could give this feature and that BitKeeper SCCS (which is clearly a reimplementation of the classical SCCS and later enhanced by features that before have been implemented by the frontend Teamware) is the distributed version control implementation that was used as a master for re-implementations like Mercurial or GIT. Schily (talk) 13:45, 22 January 2016 (UTC)[reply]
What is "Bitkeeper SCCS"? No one else calls it that, only you. I doubt if the Bitkeeper folks would even be too impressed to hear that, as it sounds a lot like taking the credit for not only a later product than anything you worked on, but even for the very innovation that distinguishes them!
As I read the later Bitkeeper release notes too, it's not even true any more that they shared a file format. Viam Ferream (talk) 13:53, 22 January 2016 (UTC)[reply]
Try to read the latest BitKeeper SCCS sources and look at the documentation. It was recently renamed to "bk", but it originally appeared as BitKeeper SCCS and both the latest available sources and available recent documentation make it obvious that this is an enhanced SCCS implementation. 14:03, 22 January 2016 (UTC)
How would I get access to Bitkeeper source, given that it's commercial and infamously protective [1] about such? Although there is this comment, "In bk-6.0 we introduced a new on-disk file format that ditches SCCS completely." Viam Ferream (talk) 14:21, 22 January 2016 (UTC)[reply]
In other words: you are unwilling to use a search engine. Well, it would have verified that your claims are wrong, the file name is: BitSCCS-0.5.2-SCCS.tgz Schily (talk) 15:45, 22 January 2016 (UTC)[reply]
[edit]

Hello fellow Wikipedians,

I have just added archive links to one external link on Distributed version control. Please take a moment to review my edit. If necessary, add {{cbignore}} after the link to keep me from modifying it. Alternatively, you can add {{nobots|deny=InternetArchiveBot}} to keep me off the page altogether. I made the following changes:

When you have finished reviewing my changes, please set the checked parameter below to true to let others know.

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—cyberbot IITalk to my owner:Online 01:49, 10 February 2016 (UTC)[reply]

[edit]

Hello fellow Wikipedians,

I have just modified 2 external links on Distributed version control. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

  • If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
  • If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 12:19, 11 September 2017 (UTC)[reply]

by Atlassian: to be included here? 95.130.167.71 (talk) 12:26, 11 March 2024 (UTC)[reply]