"Over the entire history of the project, 148 contributors have submitted code. 42 have done so in the last year."
It should be over 200. Please see:
When I check the enlistments:
I notice that this:
Step 1 of 3: Downloading source code history (Failed) tiki
It is written: "We are already aware of the problem, and you do not need to take any specific action. We usually correct failures within the next business day."
I have checked several times and the error message seems to persist.
Checkout now seems to be OK:
Ohloh update completed 13 days ago.
But stats are way off. For example:
Overalls stats seem to be missing a big chunk of CVS
Broken again as of now:
Step 1 of 3: Downloading source code history (Failed)"
I am investigating this now. The job failure itself is a bit of a puzzle, and I'm trying a fresh import now.
I'm not sure why so many of the commits seem to be missing. The CVS Monitor web site crashes when I attempt to query your list of commits, so I may be stuck drilling through rlog files for a while (ugh).
It may be a branching issue. Can you explain the branching pattern for this project? Are the missing commits on a branch other than the HEAD?
It would be very helpful if you could find a particular commit which we are missing -- preferably a relatively old one, since we haven't successfully downloaded the tiki module in a few weeks.
Thank you for your help.
We have 2 branches (BRANCH-1-9 and HEAD) at the moment. We had other branches (ex.: BRANCH-1-8) before but they are all deprecated.
Missing commits seem to be on BRANCH-1-9
Here is an example of a commit to BRANCH-1-9 in August 2006
The problem is still not solved.
Ouch, sorry I missed your post. You've been waiting a long time.
The answer to your question might simply be that this is by design: we only import a single branch. For CVS, this is always HEAD. If all of the HEAD commits are present on Ohloh, and none of the BRANCH-1-9 commits are present, then our system is working as designed.
Also, it looks like module 'tiki' is still failing to update. I'll take another look at this.
It makes sense to only import a single branch so as not to count the same effort twice (when it is merged up).
However, it skews our stats. We have some developers, such as myself, which focus on bugfixing. And some others are more focused on developing the amazing features of the next version.
The mystery is solved. That's the way the cookie crumbles :-)
Hi again Robin!
Would there be a way to use another Branch than HEAD?
It would be much more representative to use BRANCH-1-9 because it's the stable, official release. And change to BRANCH-1-10 when that comes out.
We currently don't have this capability for CVS.
To be honest, I think it would fairly straightfoward to implement -- but we just don't have the resources right now.
We're pushing to open up the Ohloh code, and this might be a good task for an outside contributor to take a swing at when the code is available.
Please inform them there is a bounty of a keg of beer (provided mail-order of beer is legal where they are) for the people that fix it.
Will the keg be delivered by the Silver Bullet train?
More likely by UPS or Fedex :-)
ok, we are now using SVN instead of CVS and it seems to have affected the stats.
169 contributors here:
(should be over 200 but at least, it's closer)
but only 47 contributors on the widget:
This is a problem of poor communication on our part.
Ohloh counts "active contributors" as contributors who made at least one commit in the last 12 months.
There were 169 total contributors found in the source control logs. However, only 47 of those contributors did any work in the last 12 months.
So in the widget and most other places, Ohloh reports the number of contributors on this project as 47.
I admit in looking over the report that this is not at all clear.
1- The widget shows the total number of lines of code and the total value of the project, so why not show the total number of contributors?
This has surely been debated/discussed.
In our information, we say that over 200 people have contributed to the source code of TikiWiki (which never ceases to impress me). This makes it awkward for us to put the nice ohloh widget, as you can see here:
2- The "Effort" of 652 Person Years seems too high to me. TikiWiki is 5 and half years old. This would be an average effort of about 120 persons per year since the beginning of the project. While we had loads of people working on the project, most are not full time. Maybe something is being counted in double?
1- Actually, I don't know that we really debated it much; this is the first time it's come up. We can debate it now. :-).
I'm inclined to agree that in the context of a 50-pixel widget with little room to explain things, "total" is probably the expectation.
I don't know whether anyone would have a cow if we switched from "active contributors" to "total contributors" in the widget. We have a lot of widgets in the wild now.
2- The person-years effort estimation is strictly an artifact of COCOMO, which I don't really defend. In my opinion, COCOMO often doesn't correlate with the typical ad-hoc development process of many open source projects (as opposed to corporate waterfall methodologies, for which it was designed). We provide it here because it's a relatively standard way to express the scale of a project to non-technical users.
Ironically, given the source control history, Ohloh could provide a very accurate estimate of actual development effort, but COCOMO is a "standard" and it doesn't use this extra information. Also, COCOMO is intended to include a lot more than just development effort -- it's supposed to estimate planning, discussion, testing, management, etc.
We can look into providing Ohloh's calculation of the total person-years from the source control logs as a new statistic.
Assuming nobody will have a cow, may I suggest a switch to "total contributors" in the widget?
Assuming it's fairly easy to implement, we can always rollback if it provokes a debate, which I doubt.
I think it's because CIA also takes into account the CVS stuff. We still have our stable BRANCH-1-9 on CVS, while the newer BRANCH-1-10 is on SVN.
I wonder if I should add the CVS to the enlistements?
But I am worried it'll mess up the stats.
Anyways... I won't do anything for now and until someone has some good advice.
I am hoping it will help with our stats. This is the branch we are really using until we release 1.10.0 (soon!)
Effort (est.) 669 Person Years
Avg. Salary $ year
seems much too high. I think something is counted in double.
I'll revert if it makes things worse :-)
I am reverting because it is now reporting:
"Short source control history (4 months)"
and only "44 contributors".
One day, I'll figure it out :-)
Recently, the enlistment below disappeared:
Yet, I can't see in the history what made it so:
I am trying to re-add but I am getting:
"The cvs server did not respond to an 'ls' command. Are the URL and module name correct?"
It looks like someone did go through and remove the CVS enlistments from Ohloh. There are a few "undone" enlistments on page 3 of the edits. (Perhaps unintuitively, undone edits are sorted by date of the original edit, which I might fix right now while we're thinking about it....)
You should be able to just click "redo" to bring them back, but I just checked and it seems we're having trouble connecting right now:
$ cvsnt -d:pserver:firstname.lastname@example.org:/cvsroot/tikiwiki ls . cvsnt [ls aborted]: Error reading from server tikiwiki.cvs.sourceforge.net: -1: Connection reset by peer
This might be some trouble on SourceForge -- we're currently seeing a lot of failed jobs coming out of SourceForge right now. This would explain why you can't add the repository using the web form.
We can try again in a few hours. For now, try simply redo-ing the enlistment edits and see if that restores the report.
I tried that and the enlistment is back.
8 days later, the CVS count has yet to succeed:
Step 1 of 3: Downloading source code history (Running 648/5470)
Maybe it's too big?