Intel to more actively engage with FreeBSD Foundation and make $250K donation
twitter.com
twitter.com
Absent volume, you and I don't get to be direct customers of Intel. I wonder who they meant. My first guess was Sony or Nintendo, since both have recent consoles with FreeBSD-derived OS, but fact checking indicated that neither have an Intel CPU.
Maybe at the behest of Apple, or the router/NAS/appliance manufacturers, many of whom run a FreeBSD-derived OS. Or perhaps it was simply effective lobbying by the FreeBSD project / foundation.
This is all speculation so if anyone knows who really prompted this I'd be interested to hear it.
(edit: grammar & fact check)
Sony - If Intel wants the next PS5 CPU to be from Intel. But Intel lacking any decent GPU, and their reluctance to low margin business may suggest otherwise.
Yahoo - some have suggested, but as much as I want to believe it is true, last time I heard they have been switching to Linux since Marissa arrival. Most of their BSD people has since left.
Netflix - FreeBSD usage is limited to their own CDN Appliance, which as the name implies, working like a appliance and needs not much tweaking. Their main, encoding and core Network is on AWS with Linux. And if Netflix is that much a big customer for Intel?
Router / NAS / Appliance? Gosh they have been here for more then a decade and only now Intel decide to listen to them? One reason could be they are switching to ARM + Linux. For them to stay on x86 they will have to better support for FreeBSD, but I still think this is a weak argument.
Apple - Despite OSX's close relationship with the BSD community. Apple doesn't use any FreeBSD on their Cloud Servers. And Airport uses NetBSD which is something different enough.
May be it is Whatsapp?
And while the FreeBSD Foundation provides governance and services around the periphery including a couple full timer contributors, almost all significant code contributions are coming from a small quorum of commercial users. If Intel follows through on higher quality and more timely HW enablement, that is going to be a lot more worthwhile than this donation.
NF was able to leverage some connections built over several people's careers to wake Intel up to the extant market share of FreeBSD users. By my naive estimation this is approaching $1B/yr.
The other conjecture I will offer is that Intel may see any of AMD Naples, ARM64, OpenPOWER as a realistic threats now and dots were connected between Intel's marketing driven culture and Intel's software services group which has a mission to make sure software runs best on Intel.
Well I hope the NF story is true. FreeBSD desperately need some love.
Uh, excuse me? They literally rewrote the whole sendfile system call to make it faster (and to make it possible to add in-kernel TLS for, well, sending files over TLS). I'm sure they would welcome, say, driver improvements for the Intel 10/40Gb Ethernet cards.
Also not saying Netflix dont love more improvement on FreeBSD. I bet they do!. But they don't seem to be key customer for Intel to do this.
Switch OS isn't FreeBSD-based.
https://twitter.com/qlutoo/status/838702899014217729
https://twitter.com/qlutoo/status/838666432774692864
https://twitter.com/ylws8/status/838777665498853376
Though the network stack may be partly taken from FreeBSD.
It's a fair bet they were considered as design candidates. Perhaps their community engagement was a factor in them losing the designs. (Though IMO that seems like it would be weighted much lower than other factors).
Speak for yourself, thank you very much.
Currently, FreeBSD supports neither Broadwell or Skylake and they're probably the most common chipsets used in laptops. source https://wiki.freebsd.org/Graphics
I would be running FreeBSD on my Dells right now if I could.
What's left for out of the box support, you ask?
* Committing the needed "linuxkpi" additions. This is complicated by the fact that some of Macy's work just imported GPL2 code rather than recreating it BSD-licensed. Hans Petter Selasky has been doing a lot of work lately to fix this gap.
* Fixing remaining bugs/incompatibilities in the linuxkpi that result in artifacts and crashes.
And the further concern is, how do we keep from falling so far behind again? Matt's vision has two pieces:
* We bring down our driver diff to almost nothing and expand the linuxkpi as needed by newer versions of the graphics drivers. This would make bringing in newer versions of the i915/radeon/amdgpu drivers less painful.
* We move the graphics drivers and linuxkpi (or portions of it) to the ports tree. FreeBSD's release cycle is far too slow to wait for a release to support new graphics hardware. On the other hand, stable branches get new ports on a regular basis.
The main area of concern for GPL in the driver space is that some of the ARM GPU drivers are GPL licensed.
From my perspective, it doesn't matter too much whether a GPL2 kernel module is in base or ports. I'm not a license purist.
Make a website where any person or company can donate a fixed amount (one-time, periodically) that gets donated to open source. The website integrates with or links to as many open source projects as possible. Perhaps the website donates to each project once a month, pooling all the money together, to keep from having lots of transaction fees.
To be more advanced, write a tool that scans all code bases and automatically tells you what open source products you use. So if I work for a company, I run the tool against all of our repositories, it determines which products we use, and it syncs with the website. Then, say we donate $100 a month, it gets split between all of them automatically in some sort of proportion.
Then github open source projects can sign up with the website and become automatically tracked with the tool, so if someone or a company uses them they can get paid.
I simply do not have the time to pursue this, but if I had the time I would. I believe it will vastly improve the open source community and help finance it.
Deciding how to split the money is the hard part here. How do you decide what each product is "worth"? Is left-pad worth as much as Apache? Probably not -- so do you say that money gets divided in proportion to lines of code? But then you reward people for writing crappy code; why write a 1000-line program if you can write a 10,000-line program and get more money? And how do you account for different levels of usage? If I have a three-line Ruby script in my repository because I didn't want to spend five minutes to rewrite it as a shell script, and I run it once a month, do you count me as "using" Ruby?
Now, a tool which enumerates all the open source code you're using could be useful. And maybe you could encourage users to make donations, and to tell you the relative value they get out of each tool. And maybe you could then suggest values based on how much other users say a particular tool is worth to them. But you're not going to get any sensible values from an automatic process; you'll need human valuation at some point in the process.
My experience is we know what code we use, management just doesn't care to pay.
(Not to knock the idea, I like it, just curious.)