Raspberry Pi - make Tier 1

..
..
..
Yes, I've been hearing about the Arm architecture moving to servers for about 10 years now. I no longer believe that it will happen, because it seems to always be forecast for "next year". In the last 10 years, all non-x86 server architectures have died one by one, and today even PowerPC is barely hanging on. Instead, we now have a healthy competition between AMD and Intel (with AMD making significant inroads), which is causing their chips to suddenly get much better. If one looks at the roadmaps for server CPUs, there has been a significant acceleration in the last 5-7 years, and it is simply because (a) AMD is trying to make an end-run around Intel, and (b) Intel is actually paying attention to improvements, because for the first time they have competition. This leap forward makes it harder for alternate server architectures (Arm, PowerPC) to make headway.

Then you need to look at AWS - https://aws.amazon.com/ec2/instance-types/a1/
I don't care at all about respect.

Well fortunate that you're not in core, then.

As a final statement, I don't believe anyone here has stated an RPi4 is a server, as in a business class/enterprise rack space server. However, as a home hobbyist server running nextcloud/nas/http/squid etc, it's a great little device.

The issue, you and other anti-RPIn people seem to miss is that a LOT of people are being exposed to Linux through these devices. It's a growing community; it's growing much faster than FreeBSD (IMHO). It's now huge in server space, pushed by the likes of Oracle over even Solaris, loved by IBM etc etc. FreeBSD? Still hanging its hat on Yahoo and Netflix? I know in my work experiences, people want to run Squid, for example, they immediately turn to any number of Linux distributions. I push for FreeBSD and I get blank stares.

Dwindling numbers of fans means you'll eventually become something like OpenVMS: a damn fine OS that no one uses.
 
....Well fortunate that you're not in core, then....

the core is not here so it does not have to read destructive posts from users whose nick begins
with e.g. `m` . So they can continue working undisturbed.
forums members can only earn respect, if they learn to respect that some forums members here have a lot more technical expertise than themselves. e.g. the one whose nick begins with 'r' made an excellent explanation , user 'm' failed to respect that(because of the lack of knowledge), that's the way it goes in forums but that doesn't change technical or market-related facts .
 
But would it be good for FreeBSD itself? I really don't know why or whether having lots of Pi users running FreeBSD should be a goal of FreeBSD.

according an interview with the Top Tux-programmer Mr. Poettering, his own community is unbearable in disussions . that community should not be tempted :) the 1st RPI Tier 1 FreeBSD -version therefore should be kept secret 😂
 
the core is not here so it does not have to read destructive posts from users whose nick begins
with e.g. `m` . So they can continue working undisturbed.
forums members can only earn respect, if they learn to respect that some forums members here have a lot more technical expertise than themselves. e.g. the one whose nick begins with 'r' made an excellent explanation , user 'm' failed to respect that(because of the lack of knowledge), that's the way it goes in forums but that doesn't change technical or market-related facts .
I find your tone obnoxious.
I therefore choose to take no further part in this 'discussion'.
 
Remember AiX and Solaris? Great OSes.

Gone. Not because they're worse than Linux, but because Linux is good enough and the kids coming from school and university only know Linux, Windows and Mac. Not because Linux is free. No serious company runs an OS without support, and Redhat support is not cheap.

With the Raspberry, FreeBSD would have a good chance to introduce herself to 25 Mio kids and hobbyists.

But an OS for 25 Mio devices has to be secure and updateable.
[/QUOTE]

I mostly agree, with the caveat that security does not really exist on any commercial systems (neither Windows nor Linux are secure - demonstrated by the constant stream of security updates) and I would address the FreeBSD security issue, within instructional environments, with regular, golden-master based, re-flashing of the microSD's (hack to your hearts content, all your hard work will disappear on Tuesday - additionally the value of a compromized RP is almost 0 and they are not attractive targets, in part because there is nothing on them that people will pay to keep and their internet connection schedules are highly variable (not reliable botnet partners)).

Your point about AiX and Solaris is spot-on. Linux is a business, that like heroin, gives away free samples (Fedora) and once hooked, updates users to Redhat. AiX and Solaris were also businesses, but it became too expensive to pay for their independent development when an adequate (the marketplace seldom pays for more than adequate) "free" alternative existed (IBM invested rather heavily in Linux, but less than AiX was costing). The only real challenge to Linux hegemony is its GNU GPL. FreeBSD has a superior licensing model that could give it a chance to flourish, but it probably needs to find a RedHat and/or Cononical that sees the business opportunity.

To ralphnsz's point that custom or traditional embedded systems are more suitable to small solutions I would offer that Linux is exploding in these markets. The cost of a standard, though scaled back Linux version, relative to the alternatives, is 50mB of memory (which costs a quarter, significantly less than a Wind River license, and noise compared to the cost of creating and maintaining a bare-metal solution). This is why Linux, or derivations of Linux, own the small systems space. The issue moving forward have to do with licensing. For performance reasons some embedded application code works better in the OS than running in user space, but the Linux rules require that any changes to the OS be freely distributed.
 
...... Linux... ... Fedora ... Solaris ...
... AiX .... ....Canonical....Windows..RedHat....Linux, or derivations of Linux...
GNU GPL...Linux....Linux...Linux..
...FreeBSD would have a good chance to introduce herself to 25 Mio ... hobbyists.....

25 million RPI-Hooligans who ask 25 million times a day the same questions :
.... all them against only 1 peaceful from Feyenoord ...

no - rejected ;)
 
according an interview with the Top Tux-programmer Mr. Poettering, his own community is unbearable in disussions . that community should not be tempted :) the 1st RPI Tier 1 FreeBSD -version therefore should be kept secret 😂

Self proclaimed top Tux programmer? AFAIC, the further FreeBSD can distance itself from that madness the better.
 
Personally i just think its a matter of exposure... I at one time all those many years ago started out on the berkeley distro's and eventually started looking and using slackware when it appeared... and to be honest over the many years moving forward in that progression actually don't even remember when i kinda forgot about the bsd's... I've always been a build it from sources type of guy but must admit in that forward progression and the onset of all the linux distro's flooding the market like many others fell prey to the misconception of the bsd's being old and really not pertinent in today's computer scene... funny as it wasn't till i ran into a couple of members here that made me stop and think and actually take a look at the bsd's again after all these years... Now that i have done that i have pretty much totally re-invested my self back into the bsd's as i can really see a lot of value it brings to the table that to many in the linux scene are unaware of as to a lot of people when you say *nix most of them say linux and then bicker over which distro's the best...

i also think that to a large part that a big piece of the current linux scene is populated with skinners and desktop people that are really mostly invested in creating a new or their own look which basically is sitting on top of lower system stuff being kept up by a smaller group... Most that i have talked to over the years that were either coming from a winblows or Mac OS seemed to fear *nix because of the loss of comfort in the desktop and how things worked...

I just think that currently theres not enough interest by people to adapt to the rpi's and its not really a matter of good or bad thing as its more of a market interest thing...

I've been doing embedded programming now for many years and now that i am back into the bsd's i doubt i will move much further forward in any linux based developments as once you start to get over some of the initial hurdles, bsd just is a better investment for anyone trying to develop in the embedded market if your worried about protecting your sources, which i can see as a growing problem for some as more and more sbc and development boards with better resources start to appear...

Hopefully others interested in the Raspberry market will start to see the light and start thinking of bsd and not be so closed...

and i agree like so many other things these days and money that Linux is unfortunately a business and not the so-called free from corporate control idea that it was all those years ago...
 
For what little it's worth, i've been using FreeBSD for nearly two decades and i would love to have just a single good modern SBC to run small FreeBSD projects on. I don't care if it's the pi4, the rockpro64 or whatever else, but there's gotta be a lot of other people out there who just wants gigglebit ethernet and a couple of USB3 ports for a tiny NAS or whatever else. ARM cores that aren't from the stone age would be a plus.
 
For what little it's worth, i've been using FreeBSD for nearly two decades and i would love to have just a single good modern SBC to run small FreeBSD projects on.

Being in the same situation, I went for a cheap x86 based Atomic PI. After a couple of easy tweaks the board has been rock solid.

Still, agree it would be good to have at least one reasonably recent non x86 well supported alternative.
 
I think a big part of the arm embedded hold back is the lack of proper tools like workable ide's at assembly level... most of these boards have evolved sofar from the old days of 8bit and 16bit atmels/microchip and the rest where from a hobby point of view everything was there in tools for hobbiests to play and learn...

There are still dev boards from places like St or Ti that have proper tools but there different then this new movement in Sbc's like the Raspberry's and Rockchip or Amlogic that basically get a big push from people looking for media devices with more capabilities then simple Android boxes...

Most of the manufactures support stuff is still clouded in privacy or involve nda's while they all try and protect their little nitch markets which is mostly in the media player areas...

For me i am more interested in a working ide where i can develop on the core board while using its resources for the plugin boards i am designing for things like motion control and some audio development... I started as a hardware and assembly guy and still prefer working down there when designing interfaces and devices and look at these newer sbc's as replacements to the older development boards ive used for years...

Thats just my thing but as much as i like the typical linux distro's i think at some point in time developers trying to still support open source but worried about protecting core maybe proprietary rights will see the advantage to the bsd type of licencing once they become more aware of it...

bsd may be old but i believe its just been missed or overlooked because a large part of new linux users made the move in part because they can build on what others have already put out in the linux world as i think the bulk of new linux users can't or aren't willing to start at the bottom and work up... all of the bsd's are less popular to that part of the linux market...
 
I'm using an AMLogic media box running Linux and it's impressive (CoreELEC on an Odroid-N2 specifically). The minimalist amount of electrical power it uses and computational power it delivers is amazing. It just runs on a little 12V/2A wall wart and even when working hard it doesn't use more than five Watts. It's amazing to me how AMLogic did that when anything x86 based would use way more power. An HTPC would probably use upwards of fifty Watts and I don't know if it would do anything that little media box can't do.
 
I think that this is a misfocused discussion. Clearly tiers are for classify the QoSupport of a CPU platform, not popularity. The problems with many arm machines are the SoC jungle, and lots of propietary blobs and poorly documented devices wandering around. Without the ability to support a well defined set of platforms consistently, there's little chance to promote ARMv8.
I think that the versatility, documentation and the features policies, make FreeBSD a better customizable system than linux in many areas, so I can be agree to certain point to ones that say that is lots of wasted potential, but you have also admit that RPI is a fancy toy at best. Also is not a good embedded system platform because poor and unoptimized I/O, inexistent integrated acquisition system and coarse timers, and even worst teaching system because is not intervention friendly (Need JTAG, breakouts, and better debug support), and there's lot of poor documented things (like the video system).
 
The problems with many arm machines are the SoC jungle, and lots of propietary blobs and poorly documented devices wandering around.

Agreed. I would even go so far as to suggest that *no* open-source operating system has ARM based processors as a tier 1 platform. For example Debian which is often seen as a fairly portable Linux distro does not provide an image that I can just shove into a perfectly good pcduino2 (Allwinner A10 Cortex-A8). No for that, I need some weird unofficial niche bespoke pcbuntu remix.

I find it a bit of a joke when you see a bunch of downloads listed such as:

Code:
RPI-B.img
BANANAPI.img
BEAGLEBONE.img
CUBIEBOARD.img
CUBIEBOARD2.img
CUBOX-HUMMINGBOARD.img
GENERICSD.img
RPI2.img
PANDABOARD.img
WANDBOARD.img
PINE64.img
PINE64-LTS.img
RPI3.img

It shows developers (such as FreeBSD) really are trying but effectively the platform is defective. It needs some sort of BIOS equivalent otherwise it is simply a waste of time to support.

Imagine if x86 did this:

Code:
Lenovo-ThinkPad-x200.iso
Lenovo-ThinkPad-x220.iso
Lenovo-ThinkPad-x1_gen3.iso
IBM-ThinkPad-T23.iso
Dell-XPS.iso
HP-Elite.iso
HP-Z300.ico
Sony-Vaio.iso
...
x 20,000

Arm is always a disappointment. It has had at least since 1987 to mature, standardise and be used properly in computers and yet before the Raspberry Pi came out the planet was absolutely deviod of open platforms using it. Especially sad for the RISCOS community because their older machines were starting to break and they literally had no alternative other than salvaging old replacement parts. This will happen again. Once the Raspberry Pi runs out of favour, you wont be able to buy an open arm platform for another 20 years.
 
XINU and Minix3 are both available on ARM, both Beaglebone. Plan 9 from Bell Labs also has a release for Raspberry Pi that is maintained.

True but "available on ARM" and yet none of them can run on my pretty standard pcduino2 (Allwinner A10 Cortex-A8). Sounds like they are prototypes at best.

Imagine if FreeBSD only ran on one x86 machine. An old ThinkPad T60 and nothing else. We wouldn't exactly call it Tier 1 support would we?
 
Nobody making money wants another IBM PC-compatible clone situation. I just assume this sort of nonsense differentiating platforms is "competitive advantage."

Driver and instruction set support is about all that you can strive for at the moment.
 
Just to clarify up front: I don't have anything against either Raspberry Pi's or against FreeBSD. Matter of fact, I have put a lot of work into using them. For example, if you see this output from my home water monitoring system:

Code:
# eqstat -d
House Equipment status
Status is from: Thu 2020/04/30 18:34:38 and is current.

Waterlevel           9674 gal Recent change 24 gal/hour χ²ₙ=0.61
Well                  Off     Last off time:  Wed 2020/04/29 00:46:21
                              Last ON period: 54:21
                              Duty cycle 0.76% over last 277 d
MainPressure        53.30 psi Recent change -3.53 psi/hour χ²ₙ=0.00
HousePressure       52.12 psi Recent change -5.19 psi/hour χ²ₙ=0.00
Flow               268.53 gal
that is data that was acquired using a pi and processed/stored on a FreeBSD server. Last night I was in my basement until 2am, soldering optocouplers to get more pump monitoring and well control onto the Pi. But the fact that I greatly enjoy using them doesn't mean that I ignore realities.

The problems with many arm machines are the SoC jungle, and lots of propietary blobs and poorly documented devices wandering around.
Exactly. Spend some time on the official raspberry pi forum (the one run by the foundation). Over and over you will see there: the Pi and its underlying SoC are not an open source platform, and are to some extent undocumented. To run it, you need binary blobs that are created by the SoC makers (typically Broadcom). A lot of the documentation of the interfaces is simply not available except under NDA, and the pi foundation does not release them. For example, a few days ago I was trying to find out exactly what the V_IL and V_IH specification for the GPIO pins is (input voltage for the pins to be considered low and high), and there is no specification of that published for the Pi 3 or 4. There are huge discussions on the forum, with users getting very upset at the foundation and its few staff engineers, and the foundation engineers getting even more upset at the users, but the spec is simply not forthcoming. Similarly, the hardware interface specs simply do not exist; the good stuff you find in Raspbian is provided by engineers working for the foundation, who have access to the NDA documentation.

People may not like it, but that is the reality. And that reality won't change, because of economics. The pi is dirt cheap (which makes it a wonderful platform, I can order a half dozen extra pi zero to experiment with for the cost of a cheap family dinner). The reason it's dirt cheap is that both manufacturing cost, and engineering / support cost are kept to a bare-bones minimum. Sure, we could have a fully documented pi, with web pages full of fine technical detail. It would triple the number of engineers working at the pi foundation. We could have extra engineers on their staff to work with FOSS implementors. Again, that would add dozens of people. Yes, companies like IBM and HP have dozens of people on their staff to do nothing but interface to Linux and other free operating systems. Yes, we could add a support system to the pi, so if something doesn't work right, you can call an 800 number and get advice. We could also harden the hardware; for example, for a few dozen dollars (and a few square inches of PC board) we could make it so the GPIO inputs survive input voltages >3.3V without blowing up the chip. But that's not what the pi is about. It's about cheap, convenient, ubiquitous, a toy for learning, and for hobbyists.

That's all very nice, but it leaves other OSes out. They have a much harder time supporting the pi than Raspbian does. And without the pi (the 400 lbs gorilla of the tiny computer marketplace), the arm world does not have critical mass.
Arm is always a disappointment. It has had at least since 1987 to mature, standardise and be used properly in computers and yet before the Raspberry Pi came out the planet was absolutely deviod of open platforms using it. Especially sad for the RISCOS community because their older machines were starting to break and they literally had no alternative other than salvaging old replacement parts. This will happen again. Once the Raspberry Pi runs out of favour, you wont be able to buy an open arm platform for another 20 years.

Intel (and AMD) is only open because of a sort of historical accident: Their CPUs have a direct lineage back to the 70s, when their predecessors (the 8086 and 8088) were already being sold. And in those days, the specification of a CPU chip was completely published, usually in 50 pages, that's all it took. Then, when IBM took up the 808x in their PCs, that got put into the long tradition of IBM publishing all the technical details too. Everyone in computing should read the "IBM 360 Theory of Operation" from the early 60s, it set the standard for good documentation. So again, there is a long tradition of fully documenting everything.

In contrast, the Arm has nearly always been deployed as a special-purpose or embedded CPU, to a large extent in the mobile / tablet market, and for embedded (near hardware). These are areas where customers use SoC or semi-custom designs, or outright use an IP core and put it on their own silicon. For this reason, there has never been a market for merchant (channel) microprocessors in Arm, and therefore also no ecosystem of documentation, platforms, and knowledge. Even the Arm server boards today (which are in use in a few places, HP and Amazon for example) are all closed custom designs.

This is not a good place for tier 1 support, given the limited amount of resources available.
 
Back
Top