OpenBSD imports nginx into tree as future apache replacement
marc.info
marc.info
I'm so used to the constant bombardment about Git (and other DVCS:es) here, that even projects using Subversion seem to be a bit "left behind", and here's all of OpenBSD doing their thing with CVS.
I guess it proves that the tools aren't everything.
Especially git has been doing a rather poor job at insulating the user from the immense complexity that the added flexibility incurs.
This is one area where our fancy new DVCS still have to learn from their ancestors.
Many OpenBSD developers are used to CVS and don't see a strong reason to change to another system. However, they are aware of the deficiencies of CVS, and OpenCVS is supposed to fix some of them. For instance, atomic commits was on the roadmap (but I don't think it is implemented yet). You can also regularly see commit messages complaining about problems with CVS.
I know some OpenBSD developers use Git locally to track their patches before committing them to the official repository.
http://lwn.net/Articles/409635/
Time frame: three years for a safe transition.
You should weight carefully whether thats worth it.
Edit: this project may not have actually happened - my apologies.
opencvs is run on some openbsd anoncvs mirrors, but is not used on our main cvs server nor used by many developers.
Sort of, it proves that a psychotic mob of minutiae obsessed sufferers of Autism are capable of incredible productivity in spite of anachronistic tooling.
Cf. people that build houses using pre-19th century technology and refuse to use modern plumbing.
tl;dr fed the troll.
I'm a coder that lives inside of Emacs and m-x shell. What about what I just said would make you believe anything other than that I consider the person to be more important than the tool?
Fuck off, and die.
Most people are unaware of the OpenBSD workflow, and cvs commits are only the final step in it. There is little "active" development done with cvs. It's for the most part used to commit already reviewed patches. Plenty of developers seem to manage their patches with other systems, including git.
That should've been your hint that I have an admiration for them and their project.
>That should've been your hint that I have an admiration for them and their project.
You called the developers a psychotic mob of Autism sufferers. A hint of admiration, you say?
The only normal/personable coder that could is probably cperciva.
OTOH, you have Theo.
shrugs Point stands.
Since adding and removing Apache modules already requires a restart of the httpd process, a good source-based build system like ports can make adding and removing modules via build flags nearly as straightforward as dynamic loading.
It says "to provide an apache replacement for base." OpenBSD has a Base system, plus it has Packages and Ports. Packages and Ports do not go through the same security audit as the Base system. Currently the version of httpd in Base is 1.3.x, which is all patched and secure, Packages/Ports has httpd 2.2.x. I suspect they mean nginx will replace httpd in the Base system, while httpd will still be in Packages/Ports.
On Linux distributions, all software is typically "equal" - everything is a package, and the only difference is what repository it comes from.
In BSD, the base system and ports are entirely distinct. The base system is a single coherent unit installed and upgraded as one piece (although the installer allows you to omit some portions of it).
The ports provide additional software packages installed and managed separately on top of the base system. The base system is maintained and patched by the core developers and officially supported by the organization. The ports are provided effectively as a convenience, and are not covered by the same security and maintenance procedures as the base system.
I guess it's there to flatten the source tree depth and probably had something to do with tiny disks of the time, shells without tab-complete or it was just a whim. In any case the convention stuck and it is now that way in all modern BSDs.