Ask stupid questions and you'll get a stupid answers, was: Technological advantages over Linux

Aryeh Friedman aryeh.friedman at
Sat Jul 25 17:18:23 UTC 2020

On Sat, Jul 25, 2020 at 12:24 PM Victor Sudakov <vas at> wrote:

> Aryeh Friedman wrote:
> > On Sat, Jul 25, 2020 at 11:24 AM Victor Sudakov <vas at> wrote:
> >
> > > Michael Watters wrote:
> > > >
> > > > On 7/24/2020 9:39 AM, Ottavio Caruso via freebsd-questions wrote:
> > > > > Why do I have to choose between Linux and FreeBSD? Why can't I have
> > > > > both? I also use NetBSD, OpenBSD, Android, occasionally Windows.
> Am I
> > > > > a traitor? Am I an infidel?
> > >
> > > > While it's fun to test out and play with different OSes it makes life
> > > > much simpler if you standardize on *one* platform.  Most
> organizations
> > > > have standards and policies about what operating systems are allowed
> on
> > > > their servers.
> > >
> > > This is very true. And when it comes to choice between Linux and
> > > FreeBSD (as a company policy), despite my love for FreeBSD and long
> time
> > > (over 20 years) experience therewith, I find there is very little I can
> > > now rationally present as arguments to choose FreeBSD over Linux.
> > >
> >
> > If stability is not a factor then you likely have no real argument
> because
> > Linux has a long history of "borrowing" to put it politely tech from
> Do you know how to present "stability" as a measureable parameter in an
> OS advocacy process? No, Debian or CentOS don't just panic or crash
> unexpectedly, if you mean that. Even the modern Windows does not.

Really simple uptime under load.   Just give them uptime stats for
different OS's.   Also show that applications that originated from
different OS's have very different ideas of stability.   Prime example is
all the crap that FreeBSD has had to import from Linux to make X work has
made it incredibly unstable (I have to reboot 3 or 4 times a day if I am
using KDE, Gnome or any other Linux originated desktop software usually due
to it slowing to a crawl for no apparent reason [has to be resource leak of
some kind]).

> If by "stability" you mean "architectural purity" then yes, I see your
> point, but it's not very valid in the eyes of a businessman.

Stability has *NOTHING* to do with architectural purity!  (it is a side
effect perhaps but not the main cause of stability) ... stability is
nothing more does it do what you expect, when you expect it (with zero side
effects) on the *FIRST* attempt and *ALL* future attempts.

> A point "you can do XXX only with FreeBSD" is a valid point. But I
> cannot find such points for the present.

Since any general purpose computer is theoretically mathematically
identical to every other general purpose computer (the only difference
being performance, capacity and what I/O devices you can connect to it)
there is no one feature of *ANY* computer/OS you can point to and say "no
other system can do X".  This is what makes computer science a science and
not voodoo.

> Root-on-ZFS and BEs are great, no doubt, but I can do without them while
> I have AWS (or any other hypervisor) cheap and fast VM snapshots.
> > > Very simple questions from the management:
> > >
> > > 1. If we decide to run FreeBSD on AWS ARM64 instances, how are you
> going
> > > to upgrade them? Via buildworld on each? Not funny.
> > >
> >
> > AWS has all kinds of problems so why use them at all instead of a cloud
> > provider that gives you a true VM and easy migrations paths.
> Please name one, with the number of features comparable to Amazon and a
> better FreeBSD support.

Just off the top of my head: RootBSD (forget their new name), Digital
Island, Go Daddy, etc. see the following mailing list thread for more info
... there are a *LOT* of such hosting firms.

If your client/employer has an issue point to them that OpenStack is based
in part on the AWS archicture and is beyond brain dead compared to other
IaaS/PaaS solutions see NASA's Inspector General's report on it where they
concluded it was a $35M complete waste of a project: ... for example the
OpenStack manual says "the worst possible disaster to occur to a cloud is
power failure" (WTF?!?!??!?!? No way should a simple power failure
completely destroy your ability to recover from a disaster [it does and I
can send you links to email threads on this])..... AWS has the same basic
flaws (the very concept of a what VM instance as defined by OpenStack was
modeled after AWS).

> >
> > >
> > > 2. What if we eventually decide to roll out docker?
> > >
> >
> > Docker is a bag on the side of a kludge that solves no real problems
> except
> > hiding a fundamentally misarchitured OS.
> Sorry, I can't say that to the management. That is, I can, but their
> reaction will be "Please go sober up and return when you are ready to
> talk shop."

Then give them stronger evidence like Linux is only the kernel and there is
no single unified architecture for how to build anything above that in
Linux (since by definition it is not Linux).   Whereis FreeBSD (and the
other traditional BSD's) are a complete system that is carefully designed
and tested as a complete unit.

To top it off Linux is less than 30 years old but FreeBSD can trace its
routes directly (if you are not concerned with the legal fine print that
says it is not allowed to) before we landed on the moon.  Therefore
FreeBSD (and the other BSD's) have been tested under fire for much longer
and are specifically designed for never fail performance (whereis Linux
never even had that goal).

You can also point to total I/O (network specifically) throughput and show
that FreeBSD consistently beats all other OS's.   For example in the
mid-90's I wrote the first single threaded web server and it blew away all
other servers on the market at the time (and still does as far I know) when
run on FreeBSD but came to a near standstill on Linux (when the
benchmarking company we hired to benchmark it said they were unable to
benchmark it at full capacity because they were not able to overload it
even using 30 bot machines to do so).

Linux has never fixed the underlying reasons why it has such poor network
throughput (just relied on Moore's Law to solve the problem) where is
FreeBSD has continued to improve in this area.

> >  If you really want to use Linux
> > then do it as a VM on a FreeBSD host to for stability reasons.
> Again, please define "stability."

See above.   Would you rather build a swimming pool on solid ground or by
putting fresh water in a inverted plastic bag in the ocean and hope it
doesn't break?

Aryeh M. Friedman, Lead Developer,

More information about the freebsd-questions mailing list