Mostly the specification was just poorly written with many ambiguities. But also the design was complex, weird, invasive and IIRC not backwards compatible with standard RISC-V.
There's no real need for LLVM, GCC or CUDA to be memory safe. POSIX libc is of course C by definition but libc's are normally extremely well tested, and it is possible to avoid libc entirely if you want.
V8 is actually a nice case for CHERI since you can easily sandbox the JIT'd code. If you were to just rewrite V8 in Rust then you wouldn't get that benefit (you can't run the borrow checker on generated assembly). I assume they have some other sandboxing methods instead though.
But in general if you think about things like V8, that's used on high performance application class consumer CPUs. It's going to be at least 10 years before anyone has one of those with CHERI (unless ARM changes its mind about Morello). I would not bet against V8 being ported to Rust before that.
As I said I think CHERI is great technology and I hope it does succeed, but it does seem like the business case for it is not as strong as it was just a few years ago.
Is it valuable enough though. Looking at Google's stats Rust has several orders of magnitude fewer memory vulnerabilities even with `unsafe` (kind of the point). If C was at that level there's no way CHERI would have ever been proposed.
There are two counter-arguments:
1. There's a lot of C/C++ code still out there. You can't rewrite it all. I'm not totally convinced by that though because, a) do you need to? Google has shown that just writing new code in Rust is very effective, and b) AI is actually pretty decent at porting from C/C++ to Rust so maybe you can?
2. CHERI also allows really strong and fine grained compartmentalisation. This is absolutely fantastic for robustness, supply chain security and so on. If you want the absolute 100% most secure code possible, then Rust + CHERI with compartmentalisation is basically the best thing you can do. (Though Rust compartmentalisation is still not actually ready yet; it's in progress though.) That's really great but I'm not sure that level of security is needed by most projects, and also I think you can get pretty good compartmentalisation (though definitely not CHERI level) by doing something like what Xous does (basically isolation with processes/virtual memory, combined with the ability to call functions in other processes; IIRC Hubris OS does something similar).
CHERI is clever tech though and it would definitely be a boon for RISC-V if it succeeds.
If you want to run CHERI code, it's true that silicon isn't easily available, but that's simply because it takes time. Various companies are working on it (Codasip, SCI, Secqai, lowRISC, etc.).
But you don't need silicon to run CHERI code. There are various emulators available that support it. There's QEMU: https://github.com/CHERI-Alliance/qemu
There's also the RISC-V Sail model, this is the latest CHERI branch: https://github.com/CHERI-Alliance/sail-riscv (unfortunately it is a bit behind upstream master, and also a bit behind the latest CHERI spec which is still evolving).
There are also a few open source chips available that implement CHERI which you can run in Verilator or an FPGA. For example cheriot-ibex https://github.com/microsoft/cheriot-ibex . This is actually a variant of CHERI for microcontrollers called CHERIoT. Long story but the plan is to merge CHERIoT back into CHERI so it is just a "profile" of CHERI.
Obviously most GUIs are not nearly that complex so immediate mode can get you quite far. Its biggest limitation is that it makes it hard to do some layouts. Your GUI layout becomes dictated by your data dependencies which is quite awkward.
Of course you can make compressed work. E.g. you fetch 66 bytes instead of 64. Hell, Intel/AMD manage to make x86 fairly fast.
But it's definitely more awkward and has costs throughout the CPU.
I would be really surprised if the lower code density is worse than the improvement due to everything being nicely aligned. Especially because Qualcomm had actual data that it isn't (if you add new instructions with the extra coding space you free up).
> Requiring C (compressed) is necessary to avoid splitting the Linux ecosystem. Chips lacking C would never be able to run binaries compiled with C.
Yes that's precisely the point of excluding it from the RVA profiles. It would mean that Linux distros don't compile code with C enabled, so chips are free to not support C and therefore can achieve higher performance (probably). And it opens 3/4 of the instruction encoding space for use by other things.
> Also the excuse given by Qualcomm - that it would make all instructions fixed length and so much easier to decode - is just wrong. RISC-V supports variable length instructions, even much longer than 32 bits, and you've just got to deal with it.
It's not wrong. RISC-V defines a mechanism by which 48/64 bit instructions might be used, but currently none are actually defined. All existing instructions are 16 or 32 bits. Without C all instructions are 32 bits. You don't have to deal with 48 bit instructions because there aren't any.
It's possible that they will add some in future, but I'm doubtful of that because a) it would be a huge pain, and b) they didn't for Vector which is where it would have been most useful.
> Just because Qualcomm acquired a company with a microarchitecture that could only deal with fixed length instructions is no reason to break the ecosystem.
Yeah it was too late to change but that doesn't mean it wasn't a mistake.
I like RISC-V (it's been my job for the last 7 years) but this is nonsense. Not everything RISC-V is good. CLIC was awful (thankfully it has been abandoned). The spec is not especially well written - the style is inconsistent due to being written by many authors, and it is waaaay too much of a textbook rather than a proper spec. (There is some ongoing work to improve this tbf.)
There's a practically unending list of undefined/implementation defined behaviours, which is great if you want to implement an ultra minimal microcontroller with 100 flops, but pretty awful otherwise.
Requiring the C (compressed) extension in the RVA profiles was definitely a mistake. The lack of true 16/64kB pages and conditional moves are probably a mistake (though fixable).
I don't know how any of these make it more robust and mature.
(But to be clear, I still think it's pretty good overall.)
> Default lock screen experience still has a needless delay of 5 seconds when entering a wrong (even blank wrong) password, even on the first attempt.
I suspect that is not KDE's fault (or Wayland's) - it's probably PAM, which by default has a 2 second delay (+/- 50%). That default is extremely difficult to change, but you can configure it. See my instructions here: https://github.com/linux-pam/linux-pam/issues/778#issuecomme...
Also if you follow that issue you can see I've been trying to convince the PAM developers to fix it (by changing it to a 0.5 second delay, which is much more tolerable and no less secure). Unfortunately they have this weird idea that users want the delay, because it lets them recompose their thoughts after getting the password wrong or something like that.
Note the shunting yard algorithm is an iterative (as opposed to recursive) version of Pratt parsing (and also precedence climbing which is virtually identical). However as normally stated it does not do proper error checking - it will accept totally invalid input.
It is. Only the default changed. Also you can press tab if someone happens to be looking over your shoulder (and your password is so obvious they can guess it from the length).
It's really really not. By default PAM has a difficult-to-disable 2ish second minimum delay for all authentication methods. However this is completely pointless for local password authentication because PAM checks password using unix_chkpwd, which has no delay. The comment I linked to is explaining that unix_chkpwd has a silly security theatre delay if you try to run it in a tty, but that's trivial to avoid.
If you want to brute force local password authentication you can just run unix_chkpwd as fast as you like. You don't need to involve PAM at all, so its 2 seconds delay achieves nothing.
It maybe does more for remote connections but I'm not sure about that either - if you want to check 10k ssh passwords per second what stops you making 10k separate connections every second? I don't think the 2 second delay helps there at all.
> Change both the config files and you can remove the delay if you want.
This is extremely complicated. See the comments in the issue for details.
You make it sound like there was a discussion where they looked at these two alternatives and chose improving sudo over using run0. Actually I just submitted a patch for this and they accepted it. I don't work for Ubuntu and I didn't even know run0 existed until now (it does sound good though; I hope they switch to that).
Yeah I would like to fix those too but sudo is the one I encounter most. Also the existence of sudo-rs meant there was less push-back. I seriously doubt the maintainers of openssh or passwd would accept this change.
I didn't actually know that Mint had enabled this by default. That would have been a useful counterpoint to the naysayers.
If you want the original behaviour you don't actually need to change the configuration - they added a patch afterwards so you can press tab and it will hide the password just for that time.
> The catalyst for Ubuntu’s change is sudo-rs
Actually it was me getting sufficiently pissed off at the 2 second delay for invalid passwords in sudo (actually PAM's fault). There's no reason for it (if you think there is look up unix_chkpwd). I tried to fix it but the PAM people have this strange idea that people like the delay. So I gave up on that and thought I may as well try fixing this other UX facepalm too. I doubt it would have happened with the original sudo (and they said as much) so it did require sudo-rs to exist.
I think this is one of the benefits of rewriting coreutils and so on in Rust - people are way more open to fixing long-standing issues. You don't get the whole "why are you overturning 46 years of tradition??" nonsense.
