Comment by rsync | Hacker News Reader

Comment by rsync | Hacker News Reader

Parent Full thread

Well, here you go.

We (rsync.net) are based entirely on the FreeBSD platform, and have been since we first offered offsite backup in 2001. Our principals have been server and desktop users since 1998.

Further, we built JohnCompanies (the first VPS provider)[1] entirely on FreeBSD and jail.

So with that disclaimer, we are compelled to point out that the high water mark of FreeBSD ... coherency ? completeness ? stability ? ... was 4.11 in 2005.

John Kozubik posted this long and detailed rundown of the situation in January of 2012:

http://lists.freebsd.org/pipermail/freebsd-hackers/2012-Janu...

http://lists.freebsd.org/pipermail/freebsd-hackers/2012-Janu...

and there was much agreement and nodding of heads. $50k over five years was committed by Kozubik to support a new development trajectory:

http://lists.freebsd.org/pipermail/freebsd-hackers/2012-Janu...

There is a good litmus test going forward. If you see 8.4-RELEASE, and see 8.x supported as "production" and not "legacy" well into the 9.x lifecycle, things are getting better.

If you see 10.2 and 11.0 and so on before you see 8.4 or 8.5 then FreeBSD is just a research project.

[1] Yes, Verio was doing that weird virtualization thing with their e10ks (or whatever it was), but in October of 2001 we were the first to make available what anyone today would recognize as a "VPS".

My view is that FreeBSD 8 is far more coherent and complete, and just as stable, as FreeBSD 4. (FreeBSD 9.0 has a few glitches related to UFS SU+J but those are fixed in 9.1.)

As for the upcoming release schedule: The next release after 9.1 should be 8.4, early in 2013. Whether FreeBSD 8.5 ever happens is an open question; we'd like to offer major branches which live for longer than our current 5 years, but we're hobbled by some contrib code which is poorly supported upstream, so my guess is that it won't happen in the 8.x branch.

That is our view as well (that 8 is more coherent and complete, stable, etc.).

The problem is that it doesn't matter unless it has a long lifecycle with lots of minor releases.

You can't deploy on the x.0 release - a bigotry that we think is deserved given the results of 5.0 and 9.0. So you need to wait for .2 or .3 or so to be confident in the release at all. But then ... if .4 is as far as it goes, you've invested years of work and hundreds (or thousands) of thousands of dollars into just one more year of lifecycle.

So 8.4 would be a good sign, and 8.5 would be great, but we need a release that lasts for a good five years and runs up to the .10 or .11 or .12...

One of the links we just posted from that big thread (the second one, I think) is the proposal John made for picking a "long term release" every few releases and really committing both time and energy to making it last long enough so that serious investment can be made in it.

That's a two way street, btw - it's not just the ability to invest in your own processes and architecture, but the ability to invest back into FreeBSD. We've had a LOT of potential feedback and contributions that could have been made throughout 6 and 7 that just never seemed worth the time and effort given how excited everyone always is about the bleeding edge ...

Interesting to hear that rsync.net is based on FreeBSD. Just from curiosity, why don't you use a linux distro? What benefits does FreeBSD offer for you compared to Linux distros. What disadvantages would you count. (I mean if you ever had a chance to make a comparison).

Our current answer is that our platform is ZFS based, and given the muddy waters with Oracle and Solaris and so on, it's more responsible to deploy ZFS on FreeBSD.

But JUST BARELY. We've had our ZFS roadmap in place since 2009, and only just finally deployed in 2012. Our intention originally was to switch to Solaris, with real support contracts and contractual commitments and so on, but then Oracle came along.

So then we had to come back to where we've been all along, since the beginning of JohnCompanies[1] - with FreeBSD.

[1] JC deployed with FreeBSD since the first VPS product rollout was based on jail, so that was that.

Why do FreeBSD users always have to 'defend' their choice. I wonder if every time someone tells me that they run some distro of Linux I should ask them why they don't use FreeBSD.

Reply on news.ycombinator.com