Organizing Open Source

friedo asks: "Projects such as Mozilla have developed sophisticated tools like Bugzilla, an automated build status page and a ridiculously cool hyperlinked, cross referenced source viewer. Would other open source developments such as the Kernel, Gnome, etc. benefit from the use of similar tools? Should we develop more organization and automation to help further open source projects?" Good thought, but such things are up to the individual projects. However, this might be a better idea for newer Open Source projects than ones that are already well established.

hmm..

[Zurk]

i dont think these tools help many open source projects..it might help the bigger ones but for most small ones, the far simpler solution is to dump the source into a tarball on a website and let people download and submit patches via email to the original author.

Organizing/tracking open-source projects

[techwatcher]

Sometime in the past few months, I was looking up Linux in one of the main search engines, and found reference to a site which specifically offered project management facilities for open-source projects. I'm sorry I didn't save the reference, but I was very impressed with the facilities at a quick glance. Some coder out there might want to replicate my search and publicise the site (I don't do code these days -- or at least, not very much!)

Bug Tracking Systems

[exa]

Bugzilla is really one of the most sophisticated bug tracking systems. I've seen it at work when I posted a bug, and it's apretty responsive and useful system. Though it might still need a bit of work. Also Debian's BTS is very good, it helps developers to track down problems in systematic way.

I'd like to see bug tracking systems evolve into something which will automate some development and communication tasks even better. Also, cross-pollenation would make these systems stronger. Ideas from both bugzilla and debian bts could be merged.

lxr (cross-referencing linux)

[dlc]

not to nitpick, but lxr [linux.no] was designed to cross-reference the linux kernel.

Some thoughts on development tools.

[Zaffle]

Speaking as a developer of a few Open Source and Closed projects, I can say that there are some utilities out there that are a Godsend to developement projects.

The most obvious and commonly used development tool (other than the compiler!) is the Makefile. Thats the first step at intergrating things, and I don't think there are many out there that think the Makefile shouldn't be used. There are also many utils which are used commonly without realizing sometimes. Patch, diff, grep, awk, etc.

the most useful utility I use is CVS. I use it for my personal projects, aswell as group projects. It is amazingly handy. With my personal single-person projects, I use it because I develop and test on many machines, and I need a central storage of my project. In the group situation, the biggest problem I've encountered is keeping all the changes together, and coordinatly work. I think almost any multi+person development project would greatly benifit from some sort of Revision Control System.

Bug tracking systems are useful, but I find only worthwhile with larger projects that have a userbase. I've found a need sometimes for a smaller PostIt(tm) Note style system for development of non-public multideveloper projects would be very useful. Somewhere you can post a note saying; Check if their is a race-condition in xyz.

Other systems like automatic build systems can be useful for very large projects. But one of the biggest things is setting all these things up. Setting up CVS is relativly easy, it takes about a day to get things all sorted, and procedures outlined and tested. But setting up BugTracking or Build Systems takes longer. Sometimes its just better to get on with the code, than to try and work with systems that may or may not be useful to your project.

I think there is a BIG opening for some kind of intergrated multideveloper system (IMDS), that has CVS, bug tracking, automatic build systems, PostIt(tm) notes, central TODO lists, etc, etc all into one easy to install, easy to setup, easy to use system. It must be something that can be setup relativly painlessly (read: package it up in .debs and .rpms), and it should be client/server based, with little required on the client side. (cvs is good, I can tell someone to install the package, and type one command to get access. The important thing is simplicity. Developers want to say, "right, lets get things sorted, I'll just go spend 5 minutes installing this IDS, and then we can get back to work." Another important thing as I mentioned is little on the client side. Developers don't want to have to change their coding style/method to work with a new system. There are already a number of Intergrated Development Enviorments (IDEs) to code and build projects with. But bug tracking systems, cvs, etc, are all seperate.

One way to describe the system would be something that you get a few coders, give them the client system, and tell them to code, and they will be able to do so and intergrate/coordinate their efforts.

---