Git User's Survey 2007

Hi all,

We would like to ask you a few questions about your use of the GIT version control system. This survey is mainly to understand who is using GIT, how and why.

The results will be discussed on the git mailing list and published to the GIT wiki at http://git.or.cz/gitwiki/GitSurvey2007

We'll close the survey in three weeks starting from 20 August 2007, on 10 September 2007.

Please devote a few minutes of your time to fill this simple questionnaire, it will help a lot the git community to understand your needs, what you like of GIT, and of course what you don't like of it.

The survey can be found here:
  http://www.survey.net.nz/survey.php?94e135ff41e871a1ea5bcda3ee1856d9
  http://tinyurl.com/26774s

I have send this survey to git mailing list, LKML, Linux-MIPS, Linux-NFS, OpenVZ, Agave, collectd, DirectFB, ELinks, Lilypond, Herrie, netconf, KDbg, Palava, OLPC, Project Sidestep, Sourcemage Linux distro, Tangram 2, Thousand Parsec, U-Boot, X.Org, xcb, D-BUS, HAL, Cairo, Mesa3D, WINE, XMMS2, XStrikeForce and LWN. Posts to X.Org, Thousand Parsec were automatically rejected: probably subscribe only lists. Posts to OLPC, KDbg, Linux-NFS, HAL, XMMS2, Sourcemage Linux distro, Elinks, collectd, netconf, Mesa3D, D-BUS, U-Boot wait for list moderator approval.

NOTE: if somebody has filled survey *before* official start, to be more 
exact before 16 August, he/she has only free-form answers saved (no 
checkbox or radio button, i.e. single-answer or multiple-answer 
questions), dues to my error when creating survey. Please fill those 
questions (and only those questions) again.

You can check how survey.net viewed your answers at
  http://www.survey.net.nz/members.php?page=single&qn=1304

Jakub Narebski <jnareb@gmail.com> writes:

Posted to linuxfr.org (french-speaking Linux and free-software news site)

http://linuxfr.org/~moy/25128.html

I couldn't find if the survey announcement made it to KDbg, collectd, netconf and itools mailing lists: all of them are "avaiting moderator approval".

I couldn't send announcement to elinks and revctrl mailing lists, as they are [true] subscribe only.

If you subscribe to those lists, could you please check if announcement made it, and if not, post "Git User's Survey 2007" anouncement? TIA.

On Aug 19 2007 01:28, Jakub Narebski wrote:

Bcc: <linux-kernel...>

What's a Bcc good for if it ain't B. :)

And http://www.survey.net.nz/results.php?94e135ff41e871a1ea5bcda3ee1856d9 Funny things there. "x86_86" arch - yep, I'd like to have that one too!

	Jan

This is partail summary of Git User's Survey 2007 after 1 week of 
running. It is based on "View Text Results" page:
  http://www.survey.net.nz/members.php?page=results&qn=1304

We have around 445 individual responses, as compared to (I think) 115 answers (Base = 115) for previous survey. That is quite a bit.

04. Which programming languages are you proficient with?

It losk like there is only 3/4 people proficient in Perl as compared to Python; it looks like Python is more popular. C is most popular, with only a few (if everything is all right with the results page) people proficient in Tcl/Tk. I'm sorry, git-gui and gitk guys; it looks like not many developers...

19. How do you obtain GIT?

Around twice as many people use binary packages as source tarball (or source package, I think). Around half the people compiles its own git (pull is also followed by compilation).

26. Which porcelains do you use?

Most people use core-git, some use cogito (the fact that it was lately deprecated and is no longer developed notwithstanding), some use StGIT, 3 even use pg (despite it is unmaintained). More people use StGIT than Guilt, but this can be cause by the fact that Guilt (formerly gq) is younger. Quite a bit use own scripts. 8 choose other... and there is no "what other" question, unfortunately...

28. Which (main) git web interface do you use for your projects?

Most use gitweb (which is distributed with git), 7 use cgit, 1 wit (Ruby), most probably XMMS2 project, 1 git-php (I wonder who...), and there are 20 "other" answers, which I am most curious about. What are they?

39. Did you participate in previous Git User's Survey?

43 people did out of 368 who answered this question, out of 115 who did participate in the previous survey. Bit curious.

55. Would commerical (paid) support from a support vendor be of interest 
to you/your organization?

Only 44 answers yes, 217 no, 126 not applicable (which was menat to encompass people who do not use git for work).

Jakub Narebski <jnareb@gmail.com> wrote:

Rather sad given their user base.  gitk seems to have at least 30%
of the respondants while git-gui is somewhere between 10-14%,
depending upon the question asked.  Odd.  qgit has a smaller
respondant base (~7.7%) but is probably the tool that more Git
users would be comfortable hacking on since it is developed in C++.
 

Hi,

On Sun, 26 Aug 2007, Shawn O. Pearce wrote:

Not only their user base, but also their portability. Tcl/Tk compile _out of the box_ on MinGW, for example. That is much more than can be said about GTK, Python amongst others.

But then, Tcl is not _that_ difficult, and we're doing pretty fine, methinks.

Ciao, Dscho

Shawn O. Pearce wrote:

I meant here that Perl to Python is around 3 to 4; more people are proficient in Python than in Perl.

It is not *that* bad, as there are 14 people proficient in Tcl/Tk, while there are 5 people who had made 5 or more commits to either gitk or git-gui. But it is not encouraging.

I wonder if PyGTK is as portable as Tcl/Tk...

Jakub Narebski wrote:

GTK does not run natively on osx, tcl/tk does.

p.s: there are some ways to make GTK run on osx, like running it using the apple x11-server, but then it looks ugly and does not have the 'native' feeling, or try to make it run with the gtk-osx port, which is incomplete :)

gabor

Gábor Farkas <gabor@nekomancer.net> writes:

Doesn't Tcl/Tk also look ugly and does not have the native feeling? But at least a user does not have to invest much work to get there.

David Kastrup wrote:

hmm.. he at least needs the X11 server installed (which is not by-default installed). but that's probably installed for a developer.

gabor

Hi,

Funny. _Even_ on this list, the git mailing list, there have been enough hints how to change the look and feel of Tcl/Tk. D'oh.

Ciao, Dscho

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

The common user experience is not magically improved by somebody having discussed the problem somewhere sometime. D'oh.

On Mon, Aug 27, 2007 at 01:28:39PM +0200, Gábor Farkas <gabor@nekomancer.net> wrote:

Maybe you're not aware of http://developer.imendio.com/projects/gtk-macosx

Mike

Mike Hommey wrote:

no, i am aware of it, i even mentioned it in my mail, in the P.S. section

for example, on this page: http://developer.imendio.com/projects/gtk-macosx/build-instructions

they write:
"
NOTE: This is mainly meant for developers wanting to help out with GTK+ 
Mac OS X, not for users. The port is not yet finished or usable for 
mainstream use.
"

gabor

Hi,

On Mon, 27 Aug 2007, Jakub Narebski wrote:

It is not. I ended my attempts with contempt after two weeks to compile _Python_ on MinGW. I came to the conclusion that the developers of Python do not care about being able to compile Python on Windows with Open Source tools only.

Of course, that means that you cannot compile Python modules with MinGW (at least as far as I know).

Ciao, Dscho

On 8/27/07, Jakub Narebski <jnareb@gmail.com> wrote:

Tk is integrated in the official Python library while PyGTK is not so I guess Tk wins even on Python (Idle [included editor] and matplotlib uses it as default).

PyGTK works pretty well though even on win32 (after installing a handful dependencies). I haven't tried to compile any of these myself though, compiling these things on win32 is usually a PITA.

//Torgil

On 8/26/07, Jakub Narebski <jnareb@gmail.com> wrote:

Ruby people, represent! (Hopefully I'm not the only one.)

All right! Someone actually uses my crappy Wit interface!

Are questions like this at all indicative of where Git is looking towards going?

Hi,

On Sun, 26 Aug 2007, Dan Chokola wrote:

No. These questions are just questions. Git is GPLv2, and it will probably stay that forever, so you are _free_ to start a commercial support scheme for Git, but others are free not to choose it.

I for one could imagine training people, and giving technical support for companies needing that support.

Ciao, Dscho

On Aug 27, 2007, at 10:48 AM, Johannes Schindelin wrote:

You seem to imply that GPLv3 would prevent any form of commercial support for Git...

Section 4 of GPLv3: "you may offer support or warranty protection for a fee."

Hi,

On Mon, 27 Aug 2007, Benoit SIGOURE wrote:

No, I was not. I was only being overly precise in my description. My apologies ;-)

Ciao, Dscho

Dan Chokola wrote:

As Dscho wrote, I should think not, but having a company basing its business around an opensource product usually means a lot of company-like features get implemented, such as various forms of reporting, documentation, integration with bugtrackers and trouble-ticket systems, etc, etc...

It is the way of companies to throw money rather than competence at temporary problems, while more altruistic organizations such as opensource projects do the exact opposite.

I for one would love if some support company could give git courses in sweden, since that'd mean I wouldn't have to. ;-)

On Mon, Aug 27, 2007, Dan Chokola wrote:

This question was actually requested on git mailing list to be in the survey...

This question is to get to know if there is sufficient demand for commercial git support for it to be viable. There are two possible outcomes:

If there is enough demand, then company selling git support can make a living. Company selling support for git means more git user's (because some corp needs to have commercial/external support to be present) and possibly more developers. It is possible that some revenue falls to main git developers, be it by paying somebody to hack on git, or paying for some feature or for fixing a bug fast.

If there is not enough demand, better to know it from onset that to have company which tries to provide git support bancrupt. No failed support company means no bad publicity on git; it _might_ mean that community support is enogh for git...

Note that commercial (paid) support does not clash with GPL. One does not need change in license to have commercial git support. GPL allows paid support.

Jakub Narebski <jnareb@gmail.com> writes:

If the few developers are efficient and responsive due to their choice of programming language, the net result might still work out fine. But it might make the the project more susceptible to discontinuation if existing developers can't sustain their involvement for some reason. On the other hand, code being inscrutable because of not being expressible well in a more common language also carries its dangers.

I found guilt essentially unusable for me due to its documentation. There is only man-page level documentation for the various commands comprising it, but the overall design is just "this is just like patch sets in ..." uh, Monotone? Don't remember which it was. Anyway, the docs were quite useless to me as someone who did _not_ previously use the system mentioned as reference.

On Mon, Aug 27, 2007 at 07:45:40AM +0200, David Kastrup wrote:

The system guilt is more like is "quilt". For people who are using pure patches to maintain changes against mainline, quilt is the "rcs" of that particular problem domain. It would probably be a good idea to have a pointer to Andreas Gruenbacher's, "How To Surive With Many Patches or Introduction to Quilt" paper.

						- Ted

On Aug 27, 2007, at 2:51 AM, Jakub Narebski wrote:

I replied "other" because I didn't know whether all questions were mandatory or not :D I don't use a git web interface... for now.

So maybe 19 other people have been as dumb as me and replied "other" for a similar reason :)

This is partial summary of Git User's Survey 2007 after 2 weeks of 
running. It is based on "View Text Results" page:
  http://www.survey.net.nz/members.php?page=results&qn=1304

The same information but in graphical form you can see at
  http://www.survey.net.nz/results.php?94e135ff41e871a1ea5bcda3ee1856d9

10. What other SCM did/do you use?

Note that this question does not distinguish between SCMs/VCSs which were used prior to Git and used no longer, SCMs which are used beside (in parallel) to Git perhaps interacting with Git, and SCMs which are used instead of Git. Also note that this is _Git User's_ survey, so it those number for example do not represent number of e.g. users of Mercurial as compared to e.g. users of Subversion.

Below there is table of SCM used, sorted by the number of responses. Note that annotations (like "a little CVS") were not weighted here.

There were 502 responses, 25 null responses (eviovalent to "none"). Only SCMs which has count more that 10 are shown. One person can (and usually did) chose more than one SCM.

  Name                      Count
  --------------------------------------------------
  Subversion                399
  CVS                       362
  Mercurial                 73
  Darcs                     59
  GNU Arch                  52
  RCS                       45
  Bazaar-NG                 40
  Perforce                  34
  ClearCase                 31
  Monotone                  23
  BitKeeper                 23
  Bazaar                    17
  SVK                       16
  SourceSafe                11
  SCCS                      10

  tla+baz+bzr               109

As you can see two most popular SCMs are Subversion ('svn') and CVS, with Subversion being slightly more popular. Among distributed SCMs with most count are Mercurial ('hg') and Arch and its descendants ('tla', 'baz', 'bzr').

35. How does GIT compare to other SCM tools you have used?

  Answer                    Count
  ------------------------------------------------
  Better                    388
  Comparable (equal)        73
  Worse                     24

  TOTAL                     485 (129 null)

Clearly Git is superior SCM! (In the minds of _Git users_) ;-) Seriously, one should take into consideration that those results are biased, because it is _Git User's_ Survey, and people usually choose SCM because they think it is best choice.

No answer (null answer) might mean that responder does not use and did not use other SCMs to compare, or at least think that he/she does not have sufficient basis for a comparison.

====================================================================

26. Which porcelains do you use?

Multiple answers (one can use more than one porcelain).

  Porcelain                   Count
  ------------------------------------------------
  core-git                    428
  Cogito (deprecated)         45

 Patch management interface layers:
 ..................................
  StGIT                       37
  Guilt (formerly gq)         13
  pg (deprecated, abandoned)  7

  my own scripts              74
  other                       11

It is understandable that Cogito still has some users, even though it is deprecated, and [I think] all of its functionality can be found in git-core porcelain. It was meant as SCM / porcelain layer when git-core didn't have it and consisted almost only of plumbing commands.

Quite a bit of people use patch management interface: StGIT, Guilt, even deprecated and abandoned pg (Patchy Git). StGIT has more users than Guilt, although that might be caused by the fact that StGIT was here longer...

It is interesting that quite a bit of responders script their git usage: 74 "my own script" users.

I am wondering what those 11 other are...

27. Which git GUI do you use?

Multiple answers (one can use more than one GUI). Note that for the first week and a bit of survey "CLI" answer had no explanation that it means command line interface, so results might be bit skewed.

  GUI                         Count
  ------------------------------------------------
  CLI (command line)          295
  gitk                        266
  git-gui                     91
  qgit                        68
  giggle                      43
  gitview                     13
  instaweb                    13
  tig                         38
  (h)gct                      3
  qct                         3
  KGit                        7
  git.el                      25
  other                       10

  giggle + gitview            56

As one can see almost as many people use gitk as CLI. Most used GUI are gitk and git-gui, most probably because they are distributed with git, and because they are portable. QGit is also quite popular, although GTK+ viewers, namely giggle and gitview have the same count summary (note that there might be instances of users using both giggle and gitview). I am a bit suprised about Giggle, I'd say.

Tig (text-mode interface for git) and git.el (GIT mode for Emacs) are also quite popular.

I wonder what are those 10 other GUI... and I didn't provide "What is this 'other GUI'?" question...

=====================================================================

44. Do you use the GIT wiki?

  233 yes, 239 no, 112 no answer

56. Do you read the mailing list? 

  168 yes, 303 no, 114 no answer

59. Do you use the IRC channel (#git on irc.freenode.net)?

  148 yes, 281 no, 198 no answer

(I do wonder a bit about "no answer" here. Does it mean: "There is git wiki/git mailing list/#git IRC channel?!? I didn't know."? Or does it mean something else: "I was too tired to answer this question...")