FreeBSD 5.0 Delayed One Year

Satai writes: "FreeBSD 5.0-RELEASE has been delayed a full year, until November of 2002. The reasons included a lack of support for SMPng - including a developer fall-off ratio of 15 to 1 - a desire to finish the PowerPC/Sparc64/IA64 architectures, and a general desire to robustly test the additions. The economic downturn even makes an appearance in the announcement."

Comments

developer fall-off

[TechnoVooDooDaddy]

As a follower/lurker to BSD-Hackers, I offered to do some device driver development, apparently not to the liking of some of the leaders on there

*cough*TIM*cough*

anyway, at least the bsd-hackers forum can be quite hostile, and i've seen it keep more than a couple people away..

Re:developer fall-off

[duffbeer703]

The policy of FreeBSD's developers is not to cater to newbies. Linux and FreeBSD are targetted towards different segments of users, why can't we just accept that? Take a look at a typical posting from a Linux user on the freebsd-newbies list. We're talking two different worlds here.
I am relatively young to the scene myself, but let's take a walk down memory lane say six years ago. Back in those days the Linux Howto's, especially the Installation Howto, were essentially Slackware Howto's. (The book I used to figure out how to install Linux was essentially the Howto's printed out.) My PC's BIOS from that era did not support booting from an ATAPI CD Rom drive. Hard drives were much smaller but the EIDE ones were coming up against a succession of limits, limits in where a kernel could be located and still be seen by a bootloader. For Linux there was a well-defined path introducing newbies: you installed and created a custom bootdisk. Linux installation instructions also told how to edit the kernel for the bootdisk floppy to change the root partition location.

From my newbie perspective, this was installation Nirvana! I didn't have to worry about LILO if I didn't want to. From the perspective of other people sharing the PC I used, other than taking up hard drive space, they didn't have to know Linux existed. And Linux could be installed in an extended partition not just a primary partition. Keep in mind that hard drives were a lot smaller then, so for dual-boot setups it was nice to be able to dedicate some more room for the Windows C: drive. And not only that but since everyone did the custom bootdisk compiling as a rite of passage, people could compile bootdisks to help others if the default floppy didn't have the right drivers.

Now from what I have read of the FreeBSD community's thoughts, they couldn't care less about such concerns. The ISP I used back then was hosted on a collection of FreeBSD boxes, abandoning a more monolothic solution with an SGI server, because the ISP's lead technical person knew how to do it. FreeBSD is more like an industrial consortium as far as the core developers go, and at least at that time there was a huge emphasis on stuff related to running ISPs. From their perspective it was laughable to devote much effort to support the most unreliable medium of all, a floppy, for custom booting a machine. And someone like an ISP wouldn't be using EIDE, they'd be using SCSI. 528MB limit, "get some real hardware, kid" I'd imagine they'd think. And they'd have their internal network and their own procedures for mass replicating setups to many machines.

Six years later I think we can see everyone got what they wanted. The Linux community developed critical mass and got wildly popular with newbies. The FreeBSD community was left alone by the newbies they didn't want to deal with.

Re:developer fall-off

[Brian Feldman]

Do you know how to write modern FreeBSD device drivers? Have you tried before? You must not know the many, many hours of work it takes for someone to learn the APIs, even if it's someone else spending HIS own hours teaching him.

It's, quite frankly, more than a little bit easier for existing FreeBSD developers to write a given driver than it is for them to teach someone what to do. Writing drivers involves intimate familiarity with the system, especially with a system where the kernel API has been in a constant state of flux in the long-running development branch.

That said, intelligent questions about an arbitrary topic with non-obvious answers are USUALLY responded to politely. You can't just say, though, "I'd like to write some drivers. Can you tell me how?" or anything even moderately like that.

Re:developer fall-off

[ethereal]

It's, quite frankly, more than a little bit easier for existing FreeBSD developers to write a given driver than it is for them to teach someone what to do.

That's an easy way to write that device driver. That's not the best way to get a whole lot of device drivers written, or to get a whole lot of new developers. Which is fine if you don't want a lot of new developers; but then people shouldn't complain when your next release gets pushed back a year :)

Re:developer fall-off

[Anonymous Coward]

I am a Professor who has been following FreeBSD and thinks that many features of FreeBSD are superior (the Man pages are the best I've ever seen, and the centralized packages/ports administration is good stuff too). However, it is not easy to teach its internals to my students, which means my research will probably wind up happening in a Linux environment due to the fact that they know it and can get stuff done there.

Writing device drivers is by FAR the most common kernel modification, and having a large number of device drivers GREATLY enhances the utility of an operating system. It is unfortunate that the FreeBSD community has limited documentation on this (although I've noticed some work in this area last time I looked). However, FreeBSD has a long way to go in providing adequate resources for someone who is knowledgeable in programming outside the kernel to learn to program in the kernel. What is needed is a tutorial in device driver writing, however, I had hoped at one point that an outsider could develop this, but the task is perhaps too daunting for anyone other than an insider who really knows what is going on. I hope the situation improves, it might lead to interesting projects for some students in kernel level programming, and more committers.

Two ways to respond to such news...

[Leimy]

You can sit and whine about how its being delayed or you can get off your duff and help.

Even taking some time to run what parts of FreeBSD 5 do exist to give some valuable feedback as to how it behaves on your system could be useful.

I am just as dissapointed as anyone else about the news but I can't help but feel motivated to lend a hand in such bad times.

I will probably try FBSD 5 this weekend and see what's what. Too bad I don't have SMP...

Re:15 to 1 ?

[imp]

15:1 is way above what can be regarded as "bad luck".

First, the 15 developers haven't departed the FreeBSD project. They are just unable to devote significant time to SNMng. There's a big difference between that and what is implied by this comment.

Actually, it is abouit right for every single free software project that I've been involved in. You get a lot of interest from people that want to see something done. Then you get about a 5 to 1 "disappearing into the woodwork" once people have begun work. Lots of people want to volunteer to help, but often times they don't have the time or fully understand what they volunteered for. You get another 3-5 to 1 attrition over the next year as people need to make money in their various fields over the next year. Or as their free time patterns change, etc.

Finally, although there's only one full time developer on SMPng, there are several people that are contributing to SMPng on an irregular basis.

So it isn't all that unusual. I'm sure many examples in the Linux world could be found as well.

Warner

Re:This isn't all bad news - quite the contrary

[Anonymous Coward]

OK, that does it. We're going with Solaris x86, I don't care what you say.

Having been a programmer for an OS Company (HP/UX), I can tell you that 12 month slips are not solely reserved for free software. About the only difference is that FreeBSD tells you up front that it's 12 months, while HP says 1 month 12 times. That being said, I'd have my company start the migration to Linux immediately.