undefined | Better HN

0 pointsvoldacar4y ago0 comments

You could just use FPGAs. They aren't fast, but your homegrown chips won't be either.

0 comments

Those are also fabricated by huge corporations. No guarantee they won't be compromised by industry or government interests.

foxfluff4y ago

True, but it's harder to compromise a fully programmable device. In essence, you'd have to somehow prevent certain kinds of programming while still permitting all the rest. How is the device to know that flipping pin 121 and 122 in a specific pattern is actually clocking bits from a storage device?

I think the web is broken (user-hostile) for the same reason. It's been made to support arbitrary programming, and now you have to play this weird cat and mouse game of trying to prevent a particular kind of programming -- that can be used against you -- without breaking all the rest of the web. Google can block your browser because they can run code in your browser. https://news.ycombinator.com/item?id=30051512

matheusmoreira4y ago

You're right. I understand now that it could be a huge step forward. Not perfect but great.

1 more reply

version_five4y ago

It does seem to some extend the higher the level of abstraction, the more potential for inappropriate controls or restrictions. More restrictions exist in apps vs OS vs cpu, or in cloud services like gdrive vs an EBS volume on AWS. So FPGAs are at least a step closer to free (or harder to secure). Even if they are under control of the manufacturer, there is a lot more flexibility and less potential for "oversight" at the gate level.

There is still the question of all the peripherals, I'd be worried about internet and video and other i/o adapters only being compatible with proprietary hardware, and impossible to use with a home programmed fpga

foxfluff4y ago

Depending on what your performance target is, you might do without adapters. Low speed USB can be bitbanged even with 8-bit attiny. RP2040 with PIO can run DVI, 720p@30Hz (though you do need to overclock the chip heavily). I don't know how to implement Ethernet with digital logic (it uses pulse amplitude modulation) but bandwidth-wise it's not too challenging. Gigabit signals to PHY are 125MHz using RGMII. I think 100Mbit Ethernet would use 25MHz but I'm not actually sure. All doable using commodity microcontrollers without an FPGA.

I could see a long road to a dystopia where these legacy protocols are deprecated and they push out devices that only speak new protocols that won't let unapproved devices talk.. but I'm not that pessimistic. I can't see the copyright mafia chasing things down so far, and they'd rail up against all the industrial users.

j / k navigate · click thread line to collapse

0 comments

matheusmoreira4y ago

Those are also fabricated by huge corporations. No guarantee they won't be compromised by industry or government interests.

foxfluff4y ago

matheusmoreira4y ago

You're right. I understand now that it could be a huge step forward. Not perfect but great.

1 more reply

version_five4y ago

foxfluff4y ago

j / k navigate · click thread line to collapse