4 ms·
"perfect is the enemy of good" "And when someone does offer something like this to the community, what do they get? Negativity." Irrelevant: nobody suggested
by nickpsecurity 11y ago
"perfect is the enemy of good"
"And when someone does offer something like this to the community, what do they get? Negativity."
Irrelevant: nobody suggested perfection. Matter of fact, all I suggested was for the project to deliver on its stated goal or articles on it modify their statements to reflect reality. Many projects accurately portray their capabilities without financial incentive at all. So, I'm sure this one can, too.
"Fully open hardware is never going to be quite as achievable as open source software because the manufacturing and distribution costs cannot be driven to zero in the same way. "
Fully open hardware has been achieved many times. It just hasn't been achieved at the desktop or laptop level. QED.
"Openness remains a minority interest, and therefore the NRE cost amortized in every unit is going to be higher. Signs of people being willing to pay for open hardware (the original designers who did the development, not the cloners who didn't) are scarce."
Now you have a real problem. You have to get the costs down or just drop the goal entirely.
"It's not even an unachievable amount of money that would be required to do a fully open CPU+GPU by paid professionals. $10m at the bottom end. Much smaller than, say, the Star Citizen kickstarter. What's unachievable is the community unity of purpose behind it."
It's actually easier than you think. All it would take are a number of academic teams, probably less than 10, using their discounted tools and grant money to each produce re-usable parts of a SOC on good process node (eg 45nm CMOS). Think USB, DDR, cache, etc. They've already produced a processor (Rocket RISC-V @ 1.4GHz) and many DSP-style chips up to 160+ cores that can be licensed. Remaining pieces are small plus SOC integration and testing with tooling.
I explained all this here:
https://news.ycombinator.com/item?id=10468534 https://news.ycombinator.com/item?id=10468534
Using academics for as many projects as possible keeps the EDA, MPW, and labor costs low. They can just public domain everything they build or cheaply license it to commercial projects to fund further development. In it for them is job/tool experience. In just a few years, we'd have every major component we need. Then, a private company (or another academic team) could put it together as SOC leveraging tools they already bought with funding from their own operations, Kickstarter, grant money, private donors, whatever. Many funding models.
So, far from an impossibility, there's multiple paths to get to the SOC's, DSP's, or FPGA's needed for open-source HW. A good chunk of it is already built, some open-sourced, and some just described in papers with enough detail to copy. There's teams all over the world with skills, tools, and funding that can be motivated to do it. So, enough excuses: start cold calling on them. :)
And things like Novena can present themselves accurately in meantime as huge, black-boxes from scheming companies that are connected together in an open-source way. Seeing it that way, though, might make people concerned with their security less likely to buy it. Re-purposed embedded boards are probably a lot safer. :)
- pjc50 11y agohuge, black-boxes from scheming companies My argument is that you (as a customer buying 1x device) can't do your own verification of the device that you buy. For example the Geisler processor ( http://www.gaisler.com/index.php/products/processors/leon3 http://www.gaisler.com/index.php/products/processors/leon3 ) looks genuinely interesting - but it doesn't seem to be a commodity part that you can actually buy off, say, Digikey. If I wanted to sow doubt on the origins of it I'd point out that the parent company Cobham PLC is a defence contractor. We're back to "scheming companies". At some point you, the end user, are going to get a small black epoxy chip package which you have been told contains a particular chip design. You can't modify or realistically inspect that object. While for software you can at least attempt your own build, Ken Thompson attacks notwithstanding.
- nickpsecurity 11y ago"At some point you, the end user, are going to get a small black epoxy chip package which you have been told contains a particular chip design. You can't modify or realistically inspect that object. While for software you can at least attempt your own build, Ken Thompson attacks notwithstanding." That's true. It's why you have to do a probabilistic approach with open HW whose behavior can be compared to spec in field. Additionally, you can make sure I/O is easy to inspect so a subset of users monitoring for suspicious behavior can catch it. It's certainly going to be a black box but easier to catch problems if open HW. Also, easier to avoid them if you retarget it with obfuscations and mask/fab vendor of your choosing. Much more difficult problem for your enemies than a total black box(s) from one source and set of vendors who are all purely profit-motivated (read: subject to bribes or contract coercion). Note: Gaisler's stuff predates Cobham IIRC, mainly funded by grant money for embedded and space applications, etc. Plus licensing what he built. He's also gives synthesizable specs and code to his customers with OSS software. Risk level of Gaisler doesn't begin to compare with Freescale, Xilinx, etc. Even post-Cobham because you can inspect the product's source and even strip what you don't trust. Additionally, your easily inspected software runs on a small, black, epoxy chip. You don't know what that software is doing any more than you know what the chip is doing. Especially if either gets hacked. My point is your arguments counter FOSS software as easily as FOSS hardware on verification side with same results as my hardware security approach. Yet, you still go for FOSS software because you think that level of assurance and verification is worthwhile along with future potential it provides. I'm making same argument for HW while citing other companies and projects that do in practice what I'm encouraging to justify it as practical. The full benefits will take time to accumulate. That's fine as long as we're on that path. :)