FreeBSD as UNIX kernel for an OS

V

vitruvius-rt

Hi, folks!

Recently I've started a company but it turns out we need a system similar to an OS that will coordinate the entire platform.

Platform that drives sequential but also parallel computing tasks, in the future maybe quantum computing tasks, but as we need to make the necessary kernel modifications we consider going with FreeBSD or to create a custom kernel only for our proprietary side.
We envision that for general computing tasks the kernel will be exposed to the computing solutions already on the market, solved with the help of LINUX-KERNEL, such as NVIDIA-CUDA, TENSTORRENT-BUDA, etc.

In such a system, LINUX will be used for already solved HPC tasks where no kernel modifications are needed.
But, where changes are need, FreeBSD will carry that load.

So, I'm considering two choices regarding what path to take:
1. Should we choose FreeBSD as the core for our new OS? Considering its flexibility in kernel modifications and the permissive licensing that allows us to keep the project closed-source due to proprietary technologies involved — similar to how Sony built Orbis OS on FreeBSD for the PlayStation, and how Apple based Darwin on BSD.

2. Or do we go with a proprietary solution?


As far as I know FreeBSD is the closest thing today to how real UNIX was — especially when it comes to clean, efficient, sequential computing.


PS. Either path In terms of execution I think we can pull it off.

Thank you,

RT
 
If you want/need access to a more broad hardware base then you have to go with Linux (or be prepared to internally support and reverse engineer what you need at significat cost)...however, if you go down the linux rabbithole then you may be encumbered by the GPL.

IMHO, BSD works better for streamlined, clean, purpose specific functions.
 
vitruvius-rt said:
So, I'm considering two choices regarding what path to take:
1. Should we choose FreeBSD as the core for our new OS?
Click to expand...
Why ask us? Strange question coming from a company because surely you don't base such decisions on (community) popularity?

The reason I consider this a weird question is because it gives me the impression that you think building against the Linux or FreeBSD kernel is the exact same thing. Fun fact: it's not. Doesn't even come close.

So... should't such decisions be based on what your actual developers think and/or can handle?

I don't think anyone can reasonably answer this question other than yourself. Also because you don't really give us any useful information to go on. Like Kent said: involved hardware can be a serious factor here.

No offense... but I get the impression that the main idea here is starting another one of those lame "Linux vs. FreeBSD" topics. Boooring. I mean... here you are quoting what companies use FreeBSD yet you leave out all those names which heavily back Linux. If you want to make comparisons at least be fair about it.

And if you really want to base this decision on popularity... maybe start a poll somewhere? ;)
 
if you don't distribute your software you don't need to publish the source (even if you use linux or other gpled source).
 
We used FreeBSD as the base for our router software in late '90s. You should do your own research once your requirements & constraints are clear. And carefully consider how much you want to keep closed source. The more you open source and (try to) upstream changes, the less you have to maintain on your own. Realize that once you start modifying the kernel, you'll likely be stuck with that OS release and will have to worry about security patches on your own. If your proprietary stuff is just in a driver, adapting to newer open source version would be far easier.
 

@ShelLuser​

I do not try to make a comparison between the two technologies in order to choose in terms of popularity or anything of the sort, but simply to query a community that can point to some perhaps missed aspects. I agree completely, nobody can fully answer this, only those involved directly with the project at hand.

@covacat
The thing is, that we might, some hardware developed solutions will be shipped with some custom kernel modifications, and we do not want that to be in the public domain. We also want to push the frontend side of things with advanced UI/UX enabled by custom hardware and such modifications are in or close to the kernel, and putting such modifications open source might lead to unwanted changes in the internal architecture, changes that might, or will cause our AI system to go in bad directions.

Tesla uses Linux for their internal kernel(center console) which is a sort of UNIX with wheels, if you want, and some custom hardware, publishing such changes do not jeopardize their product, because the meat of such project is the electric car, and infrastructure is close second with custom hardware and AI capabilities, which today is lagging in terms of complete full driving. We want to start with an AI comparable capability from the get go.
We envision a system that can drive the entire stack—not built on top of an existing OS, which has their limitations and bloatware. Something clean, new, considering a slightly different paradigm in mind, we want to use UNIX because sequential and core computing is achieved elegantly here, that's why, especially in FreeBSD.

@bakul
Makes perfect sense, noted, thank you.
 
vitruvius-rt said:

@ShelLuser​

I do not try to make a comparison between the two technologies in order to choose in terms of popularity or anything of the sort, but simply to query a community that can point to some perhaps missed aspects. I agree completely, nobody can fully answer this, only those involved directly with the project at hand.
Click to expand...
This is just me, myself, and I but I can't help ask: can you share a link with us that shows you asking the same kind of question towards the Linux community? I mean.. just to put the minds of some of us at rest? It would definitely help your credibility.

And before you ask: I am fully convinced that the mods won't mind sharing such a link. I mean.. I was the one asking for it.
 
@ShelLuser
We all know that LINUX is not the one to go according to all things debated here as the core system, due to the license statements, the choice is between to go with our own or to use FreeBSD, that's all there is. As I was not involved, even remotely with FreeBSD for a while I thought that things have gone sideways, regarding license and the project as a whole, which seems to not be the case. We will go with it as it seems that it satisfies our needs.

From your rationalization I should go to a LINUX community to ask about a UNIX Kernel that embodies free to change and to not distribute those changes, wouldn't they point me here, if not, enlighten me, where should we go?

You go where the vanilla ice cream is in the hope of getting it, and I personally think FreeBSD is close to vanilla UNIX as possible.
 
You must log in or register to reply here.
Back
Top