5:07 PM - Things we've learned this week
Obviously, releasing an operating system is a very complex process. We've made several mistakes in the process.
1. Our testing methods are not sufficient.
Several issues were found just after release. Packages for the release were not tested on another machine before it shipped. (big mistake) Several ports were broken that we didn't know about. mports was not included in 0.1-RELEASE, but rather the old method of gererating ports.
To clarify, bsd.port.mk was being used instead of bsd.mport.mk. This single difference results in most of our ports not working. There are assumptions in our ports. We could fix each port, but how do we detect the OS version? OSREL won't catch the fixes we made right after release in RELENG_0_1 and the kern.osreldate is not incremented yet.
2. The videos needed to be posted on mirrors and/or google video. The most common complaint we get in IRC is that our videos were hosted on our own site and their size. These videos were generated on a Kodak digital camera which creates quicktime videos using a very old codec which does not compress well. It is 640 x 480 video with sound. Today, I moved the install video links to google video and the demo is on a mirror with a faster pipe.
3. mports don't work. Quite a few mports were not fetchable or had other build problems. We don't have an automated build/testing system setup yet for MidnightBSD ports. There is no tinderbox/pointyhat for us. We don't have the hardware to make a massive build cluster. Thankfully, a few new users have been testing ports and reporting success/failure to us. We've fixed over 20 ports reported to us and found many others that had bugs. Several ports (particularly database ports) have been updated this week to new versions. We know these ports work in CURRENT, but testing on 0.1 is still not adequate. They *should* work with post 0.1-RELEASE ports fixes. I'm setting up an old HP Pavilion 700mhz system with a refurb Western Digital drive today. (yes a drive failed on me during this process) WD sure waits to ship a drive during an RMA.
4. We are not even close to usable yet for the average joe. The majority of issues reported to us are actually usability issues. People have trouble with the website, installation, installing ports, and other issues. One of our goals is to make things easier and we have failed in that quest to date. I promise it will get better.
We take bug reports seriously. If you have trouble doing something, using something or even surfing our website, report it to us. You can use http://bugreport.midnightbsd.org/ (preferred) or post it to our mailing list email@example.com. Subscription to the list is not currently required to post to it. This may change if we get a lot of spam, but for now it is easier to get reports. Please feel free to ask questions on this list. It's like freebsd-questions and freebsd-hackers rolled into one plus some bug reports.
5. Hardware support
We do not have a list of supported hardware for MidnightBSD. As a rule of thumb, if it works with FreeBSD 6.0, it will work in MidnightBSD. In some areas, we are near FreeBSD 6.2 hardware support. ATA and sound are current. Video card support is comparable although we have xorg 6.9 instead of 7.2. Intel gigabit network support is a little behind. Intel ICH8 boards like the intel 965LT work (i have one). ICH9 support has not been merged in yet. Wireless support is not that good. We do not yet have the FreeBSD firmware loading stuff in place so loading support for newer Intel wireless or other vendors is not there yet. My laptop has a broadcom wireless adapter (linksys) and it's working with ndis wrappers, albiet a little buggy at times.
Intel Macs do not work with MidnightBSD. Mac Pro and MacBook Pro systems are known not to work. We have not tested MacBook, iMac or Mac Mini systems with MidnightBSD.
Cross compiling for PowerPC is also not working properly for G4 based systems.
I've tested MidnightBSD on Mac Pro, Dell Precision 650 workstations, home built systems with the intel 965LT, a ThinkPad T30 laptop, a Dell poweredge 2300 (p3 xeon), HP Pavillion 700mhz celeron, a compaq with a cyrix 366mhz (i think it was) chip, an iBook G4, and a Sun Ultra 10 3d creator.
The Macs didn't work. The Dells work perfectly with all devices including SCSI working. The IBM Thinkpad does not "sleep" properly, but everything else works. The HP and Compaq work as expected. The sun system worked early on but died on me last september or so. I don't know the current state of sparc64 support. The intel board has not been tested with sound, but everything else works. I have a creative labs soundblaster audigy gamer in there (audigy 1) and it works in i386 perfectly but amd64 causes weird problems during DVD audio playback.
Remember, this is our first release and we selected 0.1 for a reason. Until MIdnightBSD hits 1.0, don't expect it to be ready for regular people.