Frans, You are confusing theoretical problems with actual problems. Again, RavenDB as a case in point. I got 27 people contributing code, all of them using the Pull Request model. I had a single instance of having to prune someone else's changes (which is what I meant by not happening very often).
On Thu, Nov 4, 2010 at 12:46 PM, Frans Bouma <[email protected]> wrote: > > a) I can pull specific commits > > yeah, like that's gonna work. Changes A, B, C and D. You can pull > 'C' but you need A and B then as well. > > > b) I am going to reject any changes that I don't like. > > c) I am going to review any code that I am pulling. > > and how many of the 'committers' will do so? > > > d) This scenario just doesn't occur all that often. > > that contradicts your claim that with github things will flourish: > as there aren't suddenly a large group of new master feed committers, the > new stuff comes from people not committing to the main trunk but to their > own branches which are pulled from. > > > e) Even if it does, I can ask you to create a specific branch just with > your > > changes. > > as this increases headaches, I am sure no-one will do this > voluntairily. You then have to keep track which local copy contains which > changes. It's not a small project, the sourcecode is huge, and the # of > tests to keep track of therefore is huge too. you can't simply test things > in a small clean-room, changes sometimes affect things all over the place. > > FB > > > > > > > On Thu, Nov 4, 2010 at 12:14 PM, Frans Bouma <[email protected]> wrote: > > > > > > > Frans, > > > There are gate keepers, and they are responsible for doing that. > > > I, for example, have about 50% chance of telling the user to fix > > his code > > > and 50% to fix the code myself. > > > Remember, we are still talking about only committers being able > to > > merge > > > code to the main repo. > > > > > > sure, but if you say, pull from my repository (as you > referred > > to > > that as a plus for git), you get my changes as well. If you commit > > your > > work, your code contains my changes as well, and thus you have to > > verify my > > code as well. This can add up if I pulled from someone out there > and > > didn't > > verify the code. > > > > FB > > > > > > > > > > > > > On Thu, Nov 4, 2010 at 11:22 AM, Frans Bouma <[email protected]> > wrote: > > > > > > > > > What I wondered, and what hasn't been debated (at least not > > this > > > 'round'), > > > is: how is code reviewing done? My experience with being an > > OSS > > > maintainer > > > is that it's likely people give you 'patches' which are not > > matching > > > how > > > things should be coded, are sometimes of poor quality or > cut > > corners > > > and you > > > have to adjust them a bit to avoid a big pile of crapcode. > > Maybe I'm > > > too > > > anal when it comes to code quality but as soon as people > are > > able to > > > easily > > > add patches to the trunk without review, it's hard to fix > > that later > > > on. > > > E.g. if a rule about 'document what you add' is in place, > how > > is > > > checked > > > that a committer indeed documented what's been added? > > > > > > With a distributed system, it's harder to verify what is > > coming from > > > where, > > > if people are updating from non-master repositories: the > > person who > > > commits > > > to the master has more to commit in that case than his own > > changes: > > > also the > > > changes he pulled from a different source. How is verified > > those are > > > in the > > > same quality? > > > > > > Or is there just 1 rule: if the tests run 'it's good > enough' > > ? > > > > > > FB > > > > > > > > > > +1 for github > > > > > > > > github is much better option than codeplex+hg (in my > > opinion) > > > > > > > > The intregated support and management of pull requests > and > > the > > > whole > > > > community aspect around the source code repositories is a > > real > > > boost to > > > > contribution and easy managment. If you really prefer > > mercurial > > > than go > > > with > > > > bitbucket and you get some of the good this github > brings. > > > > > > > > From a pure capability and tooling perspective I think > > Mercurial > > is > > > a > > > little > > > > better (at least on windows) but I think that is > > compensated by > > the > > > how > > > > great github. > > > > > > > > just my 2 cents. > > > > /Torkel > > > > > > > > On 3 Nov, 13:36, Fabio Maulo <[email protected]> > wrote: > > > > > sorry NUnit is in launchpad. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > On Wed, Nov 3, 2010 at 9:32 AM, Fabio Maulo > > > <[email protected]> > > > wrote: > > > > > > Moq svn (Google code) > > > > > > NServiceBus svn + git (Source forge + GitHub) <== > > Official in > > > SVN > > > > > > Rhino.ServiceBus git (GitHub) uNhAddIns Hg (Google > > code) > > > > > > SharpTestsEx Hg (Code Plex) ConfORM Hg (Google Code) > > NUnit svn > > > > > > (SourceForge) Castle git (GitHub) Spring svn (custom) > > > NHibernate svn > > > > > > (SourceForge) NHibernate.Validator svn (SourceForge) > > > > > > NHibernate.Spatial svn (SourceForge) > NHibernate.Search > > svn > > > > > > (SourceForge) NewtonJson svn (CodePlex) Log4Net svn > > (Apache) > > > Lucene > > > > > > svn (Apache) Re-Linq svn (custom + CodePlex only for > > deploy) > > > ANTLR > > > > > > svn (Custom) SharpMap svn (Code Plex) > > > > > > > > > > > On Wed, Nov 3, 2010 at 9:11 AM, Ayende Rahien > > > <[email protected]> > > > wrote: > > > > > > > > > > >> I mean popular in the sense that out of the .NET > > projects > > that > > > I > > > > > >> follow that uses DVCS, most use Git. > > > > > > > > > > >> On Wed, Nov 3, 2010 at 1:54 PM, Fabio Maulo > > > <[email protected]> > > > > wrote: > > > > > > > > > > >>> I know lot of OSS using SVN more than Git... btw.. > > > > > >>> Mercurial is supported by Bitbucket, Google Code, > > > SourceForge, > > > > > >>> CodePlex Please give me a more detailed definition > of > > > "popular" > > > > > >>> since its translation in Italian and in Spanish can > > be > > > interpreted > > > > > >>> as "not used only by elite". > > > > > > > > > > >>> On Wed, Nov 3, 2010 at 8:42 AM, Ayende Rahien > > > <[email protected]> > > > > wrote: > > > > > > > > > > >>>> Diego, > > > > > >>>> I know of a LOT of OSS projects which are using > Git > > I know > > > of > > > > > >>>> very few using HG. > > > > > > > > > > >>>> On Wed, Nov 3, 2010 at 1:37 PM, Diego Mijelshon > > > > > >>>> <[email protected] > > > > > >>>> > wrote: > > > > > > > > > > >>>>> Oren, > > > > > > > > > > >>>>> What stats do you use to say Git is more popular? > > > > > >>>>> I think they both have lots of followers. > > > > > > > > > > >>>>> I'm slightly biased towards HG, because it has an > > easier > > > > > >>>>> learning curve and it doesn't have the > "non-native" > > feel > > of > > > Git on > > > > Windows. > > > > > >>>>> Regarding the specific points you mentioned: > aren't > > those > > > just > > > > > >>>>> Github features whose current implementation you > > like > > > instead of > > > > > >>>>> Hg/Git differences? > > > > > > > > > > >>>>> In any case, here's my 2c regarding source: > > > > > > > > > > >>>>> - There seems to be a consensus to move to a > > DVCS. > > > Nobody > > > > > >>>>> wants to > > > > > >>>>> stay with SVN > > > > > >>>>> - As others said, after the release might be a > > good > > time > > > to > > > > > >>>>> do the > > > > > >>>>> move > > > > > >>>>> - IMO, the decision should be done first by > the > > > committers > > > > > >>>>> and, if > > > > > >>>>> there isn't a clear winner, by the > contributors. > > How > > > about a > > > > poll? > > > > > > > > > > >>>>> And regarding the site (in no particular order): > > > > > > > > > > >>>>> - I've said it before: the current state of NH > > identity > > > is > > > > > >>>>> just > > > > > >>>>> terrible. Searches for "nhibernate" > "nhibernate > > source" > > > > > >>>>> "nhibernate bug > > > > > >>>>> tracker" "nhibernate docs" and "nhibernate > > binaries" > > > should > > > > > >>>>> all point to a > > > > > >>>>> unified site. > > > > > >>>>> - I don't have anything against Jira per se, > but > > having > > > it > > > > > >>>>> redirect > > > > > >>>>> to an IP is just unprofessional > > > > > >>>>> - SourceForge still feels like 1999. I really > > like > > > Google > > > > > >>>>> Code, I > > > > > >>>>> dislike CodePlex, and GitHub is meh. But the > > decision > > > should > > > > > >>>>> be made by > > > > > >>>>> those in charge of maintaining it. > > > > > > > > > > >>>>> Diego > > > > > > > > > > >>>>> On Wed, Nov 3, 2010 at 08:03, Ayende Rahien > > > <[email protected]> > > > > wrote: > > > > > > > > > > >>>>>> Frans, > > > > > >>>>>> Git is more popular than hg. And we aren't > > considering > > > > > >>>>>> centralized SCM > > > > > > > > > > >>>>>> And yes, there is a HUGE difference between > > sending a > > > patch and > > > > > >>>>>> sending a pull request. > > > > > > > > > > >>>>>> a) it is *significantly* easier to handle a pull > > request, > > > > > >>>>>> because it is a single command, rather than a > set > > of > > > operations > > > > > >>>>>> b) it allows you to have your own fork and > easily > > merge > > > future > > > > > >>>>>> changes > > > > > >>>>>> c) it means that Joe can pull from you, not just > > from the > > > > > >>>>>> master feed > > > > > > > > > > >>>>>> On Wed, Nov 3, 2010 at 12:54 PM, Frans Bouma > > <[email protected]> > > > wrote: > > > > > > > > > > >>>>>>> > I actually do have a problem with hg. I think > > that Git > > > is: > > > > > >>>>>>> > a) more popular > > > > > > > > > > >>>>>>> than what, subversion? Perforce? CVS? > > > > > > > > > > >>>>>>> > b) GitHub has tremendous pull in terms of > > encouraging > > > > > >>>>>>> contributions. > > > > > >>>>>>> > c) I saw a huge spike in the amount of people > > > contributing > > > > > >>>>>>> > once I > > > > > >>>>>>> moved to > > > > > >>>>>>> > github. > > > > > > > > > > >>>>>>> I have a hard time believing that the > scc > > system > > > used > > > > > >>>>>>> is of any relevance whether a developer is > > capable of > > > > > >>>>>>> contributing any code. I > > > > > >>>>>>> mean: > > > > > >>>>>>> it's not as if someone who changes some code in > > his own > > > branch > > > > > >>>>>>> is suddenly able to commit those changes as > well: > > the > > > change > > > > > >>>>>>> has to be reviewed, tested, agreed upon and > then > > it's > > > > > >>>>>>> committed. A svn patch is just as simple for > that > > than > > > any > > > > > >>>>>>> other patch. > > > > > > > > > > >>>>>>> I don't deny what you saw on ravendb > > stuff, I > > just > > > find > > > > > >>>>>>> it a 'coincidence' rather than a correlated > > event. > > > > > > > > > > >>>>>>> FB > > > > > > > > > > >>>>>>> > On Wed, Nov 3, 2010 at 12:31 PM, Fabio Maulo > > > > > >>>>>>> > <[email protected]> > > > > > >>>>>>> wrote: > > > > > > > > > > >>>>>>> > And move the code in CodePlex... > > > > > > > > > > >>>>>>> > -- > > > > > >>>>>>> > Fabio Maulo > > > > > > > > > > >>>>>>> > El 02/11/2010, a las 16:38, Jorge > > > <[email protected]> > > > > > >>>>>>> escribió: > > > > > > > > > > >>>>>>> > > Hello there, > > > > > > > > > > >>>>>>> > > I am in the process of downloading > the > > code > > via > > > SVN, > > > > > >>>>>>> > and it > > > > > >>>>>>> is > > > > > >>>>>>> > taking > > > > > >>>>>>> > > a very long time. > > > > > > > > > > >>>>>>> > > Can someone please enable Git repo in > > > sourceforge, > > > > > >>>>>>> > or > > > > > >>>>>>> better yet, > > > > > >>>>>>> > move > > > > > >>>>>>> > > code to Github? > > > > > > > > > > >>>>>>> > > Respectfully yours, > > > > > >>>>>>> > > Jorge > > > > > > > > > > >>> -- > > > > > >>> Fabio Maulo > > > > > > > > > > > -- > > > > > > Fabio Maulo > > > > > > > > > > -- > > > > > Fabio Maulo > > > > > > > > > > > > > > > > > > > >
