Hacker News Reader

Comment by tiffanyh | Hacker News Reader

Love FreeBSD but ...

Really wish there wasn't a split between FreeBSD & Matt Dillon 20-years ago, since DragonflyBSD is so strong and yet FreeBSD hasn't benefited from it's innovations.

They'd truly would be better together than apart.

Can only imagine how much stronger the BSD ecosystem would be overall today, if the past 20-years these teams had been together.

https://www.dragonflybsd.org

Announcement:

https://lists.freebsd.org/pipermail/freebsd-current/2003-Jul...

EDIT: "stronger together" meaning, FreeBSD would benefit from the technology innovations, architecture simplicity and performance that DragonflyBSD brings ... and DragonflyBSD benefiting from the larger install base that FreeBSD has (and it's brand recognition).

View on HN

Complexity management is key in any project. Lots of successful forks reduce total complexity. NetBSD and OpenBSD are excellent examples — one could argue that their goals and innovations (in portability and security) would also be useful in the «mainline» FreeBSD, but that would explode complexity.

The FreeBSD community is carefully watching progress across many forks and cherry-picking / importing a lot of the cream.

As an outsider, I've actually seen the most engagement between OpenBSD & DragonflyBSD.

E.g., there's active exploration of bringing the Hammer2 filesystem from DragonflyBSD to OpenBSD.

https://github.com/kusumi/openbsd_hammer2

I’m not sure what to make of it, but amusingly your link says “OpenBSD version of netbsd_hammer2”, which itself says “NetBSD version of freebsd_hammer2”, which in turn says “FreeBSD version of openbsd_hammer2”.

They’re all under the same profile so I think everyone is sharing?

I also would prefer that this split would not had place and that DragonflyBSD would never start ... but yet it happened ... so what?

Recently (as a seasoned FreeBSD user/sysadmin) I tried DragonflyBSD to check how it went ... and it lacks a LOT of things that FreeBSD has - like for example the GEOM framework.

On FreeBSD its quite easy to list block devices like:

    FreeBSD # geom disk list

... but on DragonflyBSD its not possible as there is no GEOM ... so the only thing you can do is to:

    DFBSD # camcontrol devlist

... but its far from usable and its very limited.

On FreeBSD I was able to write lsblk(8) Linux replacement ... I would not be able to do that on DragonflyBSD.

Yes - its a pity that such a talented person such as Matt Dillon left FreeBSD project - but DragonflyBSD is not a better replacement also ...

Regards, vermaden

I think it's fair to say that DragonFlyBSD has diverged with vastly different goals to FreeBSD. It's a good thing it's able to do what it sets out to do. It's not trying to replace FreeBSD.

It's been a long time, but IIRC the split was due to direction on SMP. From what I recall, the FreeBSD team made the right call at the time.

https://people.freebsd.org/~kris/scaling/dfly.html

> “ From what I recall, the FreeBSD team made the right call at the time.”

I’m not so sure.

Given how unstable FreeBSD 5 (intro of new SMP) was, until it finally stabilized multiple years later at v8.0 - I’m not sure others would have that same sentiment.

And seeing how DragonflyBSD performance is standing up to FreeBSD (and best’ing it times) even though they have a radically smaller developer community (and no corporate support) - it does challenge the topic of who was right.

I do think objectively, people would state that DragonflyBSD chose a clean/simpler architecture that’s easier to understand

https://www.phoronix.com/review/bsd-linux-eo2021

FreeBSD 5.0-RELEASE was announced January 19, 2003.

Matt Dillon announced DragonFly BSD nearly 20 years ago on July,16 2003

FreeBSD 8.0-RELEASE was announced Nov 25, 2009

You’re talking about events that took place over 13.5 years ago.

On the benchmarks:

Where are the Apache, NginX, PostgreSQL and Redis benchmarks?

Where are the measurements of network performance?

>Where are the Apache, NginX, PostgreSQL and Redis benchmarks?

>Where are the measurements of network performance?

You'd have to ask Larabel about that. But given that Dragonfly can seemingly trade blows with FreeBSD *18 years later* does seem to suggest there's some merits to their design

DragonflyBSD design is definitely not without merits.

But FreeBSD took the mainstream approach.

Here are some benchmarks,

https://www.dragonflybsd.org/performance/

Which are a decade old?

Oops, I was actually replying to the parent post about benchmarks from 2008.

https://www.dragonflybsd.org/performance/ Has one from 2012 and an update 2018 with a Threadripper 2990WX, though there is no comparison chart..

Oh.. here is a benchmark [0] from 2021, Dragonfly performs comparably with Ubuntu and FreeBSD 13 on an intel core i9 of some type. I have no idea whether these benchmarks reflect real world performance. I also don't know where and how much Dragonfly is used in 'production'.

[0] https://www.phoronix.com/review/corei9-freebsd13-dfly6

It was also a clash of personalities.

It was all truly unfortunate. But sometimes there are no clear right answers - but you have to make a choice between multiple bad options. When doing nothing isn't an option either then you have to make a call as to what you think the least terrible outcome will be and history will be the judge as to whether you were right.

Similar issues contributed to there being multiple *BSDs. It's never that simple of course, but it was certainly a part of it. Technical/ideology and Personality all play a part.

Is there any (ideally charitable, historical) writing about this for the uninitiated?

You've gotta be careful giving vikings a commit bit.

Your best bet for information is probably the mailing list archives. To be a bit reductive, Matt had strong opinions about SMP and threading and got his commit bit yanked. As a result he forked FreeBSD (4?) to create DragonFlyBSD. It's not so far off of how we ended up with OpenBSD really. Or as Linus put it, Theo is… difficult.

Honestly, I think these forced forks are good. They're a a perfectly natural result of getting a bunch of smart, opinionated people in a group together. They've given some incredibly smart people a much bigger sandbox to play in. As a result we've got things like openssh and hammer.

20 years later, who made the right call?

That makes me feel old. I was running FreeBSD back when that split happened, I can't believe it's been 20 years.

I think the BSD ecosystem was harmed more by the SCO lawsuit, and the rise of Linux than by the Dragonfly/FreeBSD split.

You might be confusing SCO and USL.

SCO brought lawsuits against IBM (Linux) in 2003.

https://en.m.wikipedia.org/wiki/SCO_Group,_Inc._v._Internati....

USL brought lawsuit against BSD in 1992.

https://en.m.wikipedia.org/wiki/UNIX_System_Laboratories,_In....

Novell bought USL in 1993 and sold the corpse to SCO in 1995. They're related entities.

You're indeed right.

Not necessarily in disagreement, but some points to consider:

* The nice thing about free software is that FreeBSD can benefit from Dragonfly's innovations!

* Dragonfly certainly would benefit from FreeBSD's (relatively) large userbase; but it could also be hurt, because FreeBSD's larger userbase would be more demanding of the kind of stability that could slow down innovation.

* ...and, likewise, FreeBSD might suffer if all those innovations harm the OS' reputation for robustness and stability.

I have to say FreeBSD itself is also totally fine on the desktop. You don't need DragonFly for that. FreeBSD is my daily driver.

Reply on news.ycombinator.com