<GenTooMan>
palmer1, hmm so if I read your concern it's that the last few years of RISC-V have been no forward spec ratification. Despite the fact that they published specs they are solid. So things are left in a quantum state being most uncertain. Companies are left rolling the dice and gambling they are applying a standard that's actually used.
<xentrac>
sounded like it was more a question of "we say stable but we don't mean what people think stable means"
<palmer1>
it's both ;)
<palmer1>
if things had progressed to schedule at least we'd have that to fall back on, but they haven't
hendursaga has quit [Quit: hendursaga]
hendursaga has joined #riscv
<palmer1>
SOC vendors end up either trying to navigate the vague committments and try to build something, or just go back to Arm
nvmd has quit [Quit: Later, nerds.]
<GenTooMan>
it kind of sounds people have been pontificating but not actually doing useful. all talk without commitment is a deadly game to play.
* GenTooMan
makes "doing useful" to "doing something useful" heh
<dh`>
well, ultimately it's what ends up in silicon that matters
freakazoid12345 has quit [Read error: Connection reset by peer]
<dh`>
since software implementations are easily changed and specs by themselves are shouting in a vacuum.
<jrtc27>
which is why you'd better make damn sure what gets implemented is a good idea
freakazoid333 has joined #riscv
<jrtc27>
otherwise you have to support that crap, or deal with the fragmentation, forever
<xentrac>
yup, there's no perfect solution
<pierce>
That's a good point, XuanTie may end up shipping the drafting spec indefinitely. They may have already invested a lot of time and money into the software around it
<pierce>
What is there incentive at the moment to go to the "actual stable" revision
<jrtc27>
it's been made very clear to them that, if they want to claim compatibility with any profiles beyond the retroactively-created 2020 (2021?) profile, they must switch to the standard spec
<pierce>
Yeah, that's if they care to support a potentially smaller fragment of the fragmentation
<pierce>
They got their foot in the door first
<jrtc27>
I imagine RHEL's baseline will be for a newer profile, given it has much stricter platform requirements than the likes of Debian/Fedora/Ubuntu
<jrtc27>
and therefore if they ever want to sell a processor that can run RHEL or its derivatives they'll have to
stikonas has quit [Remote host closed the connection]
<jrtc27>
Debian/Fedora/Ubuntu being "we'll try and support everything within reason", but RHEL being "if you're not a boring ACPI + 'server/workstation-class' hardware you can get lost"
<pierce>
That's fair. I don't know what XuanTie is doing, but maybe they don't care?
<pierce>
It's pretty much a doomsday scenario
<pierce>
All in all, it's unfortunate that it was called stable in respect to the revision of the spec, rather than the entire spec itself
<palmer1>
IMO we're going to have v-0.7.1 in everything, for a long time
<palmer1>
There's really no way around it at this point, it's just going to be a huge headache
<jrtc27>
the next 5 or so will be painful
<pierce>
I think we'll find out when a Nezha D2 exists
<jrtc27>
in 10 years time nobody will care about some crappy old severely-underpowered SoC like the D1
<palmer1>
They've already said they're sticking with their fork for now
<pierce>
Fork of...?
<jrtc27>
instruction set
<pierce>
It's not their for though is it?
<jrtc27>
in userspace you can ignore it as a brownfield custom extension that you forget exists
<jrtc27>
in the kernel you can't because they deliberately went and violated the privileged spec
<pierce>
It's just the "stable (but not really) pre-stable" spec plus extra garnishes
<pierce>
s/their/theirs/, s/for//
<jrtc27>
it is now
<jrtc27>
it's being branded as Xtheadv
<jrtc27>
or something like that
<pierce>
Oh I didn't know about the privileged spec changes
<jrtc27>
to stop them lying about it being RVV
<pierce>
jrtc27: Uh oh
<jrtc27>
because all their marketing claims it's V, which it's not, and I reported to whichever crowd-funding site they were using but nobody seemed to care all that much
<pierce>
It's not a lie
<jrtc27>
apparently "here's evidence that you're lying" isn't enough
<jrtc27>
it is
<jrtc27>
V means the ratified V spec
<jrtc27>
that does not exist
<jrtc27>
therefore they are lying
<pierce>
It's like arguing that a baby is not a man, but instead a boy
<jrtc27>
doesn't mean they're actively trying to deceive, but it has that effect
<jrtc27>
uh, no
<pierce>
It's baby V
<GenTooMan>
so it's sort of like saying that "this device that ISN'T A V spec" but we say its kind of the V spec? but not at all?
<palmer1>
Well, it's hard to blame the Allwinner folks for calling it a V implementation when foundation keeps doing that too
<jrtc27>
yeah I do not like how the foundation is supporting them
<palmer1>
IMO that's the real problem here
<jrtc27>
saying "our messaging was confusing so we'll be lenient on you" is one thing, promoting it is another
<palmer1>
It's expected to have vendors try to play these tricks, it's not OK for the foundation to do so
<jrtc27>
yeah
<palmer1>
Well, now they're just saying their messaging wasn't confusing, while directly contradicting that messaging
<jrtc27>
I just ignore the messaging, the story changes within single meetings, like during one of the sessions at LPC
<jrtc27>
but I only have to deal with the psABI
<jrtc27>
where it's backwards from ISA things; get software to agree on what to do then write it down, not the other way round
<palmer1>
Well, unfortunately it's really hard to do anything when every statement contradicts the last
<pierce>
jrtc27: This sounds like they haven't solidified their stance on whatever you're describing
<palmer1>
I agree the messaging doesn't really effect the psABI, we've got put a pretty strong "never break anything" stake in the ground there for that reason ;)
<jrtc27>
yeah
<jrtc27>
and I am *very* conservative about accepting new things
<jrtc27>
in part because I then have to deal with it downstream :)
<jrtc27>
I honestly don't know what would happen if during the psABI review someone came back with feedback that would be a breaking change...
<jrtc27>
'cause that would be "sod off that ship's sailed" for pretty much everything
<jrtc27>
seems like a pointless exercise
<jrtc27>
but oh well
<jrtc27>
maybe someone wants to add some commas
<pierce>
Notes for the next revision?
<xentrac>
heh
<jrtc27>
pierce: there are lots of problems; people at the top have high-level goals but no clue about what's actually happening, and people at the bottom mostly know what they're writing but often lack any clue about what the overarching policies are supposed to be and just make things up
<pierce>
That could lead to spec churn, and no one would be thrilled to have so many versions out in the wild
<palmer1>
With how the ISA is going we'll probably end up with an ABI spin at some point, just to bound the chaos
<jrtc27>
and the people at the bottom also vary heavily in general competence
<xentrac>
I wanna hack baby V
<pierce>
I wish I had the insight you guys have
<jrtc27>
palmer1: standardise CHERI-RISC-V and use that as an opportunity for a fresh start :)
<jrtc27>
that's a whole new ABI
<jrtc27>
(though so far closely tracks the plain RISC-V one, but does at least give you a line in the sand for a new baseline)
<jrtc27>
pierce: ABI-breaking changes aren't revisions, they're new ABIs
<xentrac>
so far working security is not a selling point because customers can't tell if it works or not
<jrtc27>
you *can* make them in very special cases but it takes years to transition over and needs a compatibility story in the interim
<xentrac>
maybe that has changed in the last two years
<jrtc27>
meh, consumers will buy security if you market it sufficiently
<jrtc27>
but it's not as shiny as features
<xentrac>
they'll buy fake security just as eagerly as working security if you market it the same
<xentrac>
and fake security is cheaper to add
<jrtc27>
yeah, that's why you need branding they recognise
<jrtc27>
the Intel Inside of security
<jrtc27>
sell the name not the promise of security
<xentrac>
it's almost totally random whether the name that ends up denoting working security is more prestigious than the names that denote fake security
<xentrac>
unless the consumers can tell the difference
<xentrac>
DARPA reviewers, yes
<xentrac>
PCI DSS, sometimes
<jrtc27>
well you couple it with campaigns from the important vendors
<xentrac>
sure, if you can decide what other people do, then you can solve the problem :)
<xentrac>
but what incentive can you offer Apple, Google, and Microsoft?
<jrtc27>
they want the tech
<jrtc27>
that's their incentive
<palmer1>
actually, our incentive is beer
<xentrac>
because you can be very sure that right now the DoJ is offering them incentives to not ship real security to customers
<jrtc27>
palmer1: you have to pay for beer? then why does everyone say "free as in beer"...
<xentrac>
I don't think you can realistically tell Apple, Google, and Microsoft "you can only build your servers on CHERI-RISC-V if you also use it in all your consumer devices and give the consumers the bootloader keys"
<xentrac>
so I don't think you can offer them an incentive to sell real security instead of fake security
<palmer1>
we're WFH now, so we pay for beer
<palmer1>
it's kind of a bummer
<xentrac>
they might adopt it (although even OpenTitan isn't CHERI-RISC-V, is it?) but that doesn't mean their branding will tell people what to buy
<xentrac>
it doesn't mean they'll use their brands to influence other people to adopt real security
<jrtc27>
no but if, say, you have Google promoting their Pixel phones as having CHERI, and Microsoft's Surface devices, and giving discounts or whatever to OEMs that adopt it, then suddenly you have this consumer awareness
<xentrac>
maybe, but that doesn't mean it'll be deployed in a way that gives the users more security instead of less
<xentrac>
lots of phones are running capability-secure versions of L4 apparently? but that doesn't protect Chinese teenagers from having their faces scanned by Tencent games in order to enforce the new video-game law
<xentrac>
because it's being used to protect the baseband firmware from the users, not vice versa
<pierce>
We need a cute mascot for CHERI to help market it, I'm thinking a platypus with a devil onesie and a pitchfork
<jrtc27>
what's your point? we should abandon hardware security because it can be used against the users?
<xentrac>
nope, not at all
<xentrac>
I think hardware security is very important
<xentrac>
but fundamentally a public who doesn't understand infosec is kind of doomed
<jrtc27>
yes, one of the things that does make me sad about hardware security is that it might stifle various homebrew communities, but overall that's a price worth paying IMO for thwarting classes of attacks in one fell swoop that have plagued the industry for decades
<jrtc27>
yeah
<jrtc27>
it'll get dumbed down and people will just see "security"
snowrichard has joined #riscv
<xentrac>
hardware security will be a big help to people who do understand infosec
<jrtc27>
because "memory safety" is meaningless to them
<xentrac>
I don't know if it'll turn out to be good or bad
<xentrac>
but it'll certainly be important
<jrtc27>
it could never turn out, but I hope not, and I imagine a lot of UK government people hope not, given the amount of money being poured in...
<jrtc27>
pierce: we have struggled to come up with a logo for years...
<jrtc27>
a lone cherry just ends up looking too much like an apple
<xentrac>
"cher" is French for "expensive"
<jrtc27>
and cheri is dear (in the love sense)
<xentrac>
right
<xentrac>
you could use diamonds. they're both expensive and hard to cut through. or cubic zirconia, which is additionally hard to break
<xentrac>
good thing I'm not in marketing
<xentrac>
because these are terrible ideasa
<jrtc27>
heh
<jrtc27>
or tiny grains of sand to represent the fine-grained protection
<jrtc27>
:P
<GenTooMan>
as sands flows through the hour glass so go the days of our lives ...
<jrtc27>
:)
<xentrac>
someone said that of all the doctors in the Civil War, the one who made the greatest contribution to advancing medical science was Dr. Gatling
<xentrac>
because he invented the machine gun
<jrtc27>
as in forcing future generations to have to deal with more gun wounds?
<xentrac>
exactly
<xentrac>
current generations too
<GenTooMan>
it did lead to a more forceful way of making people deal with things instead of shoot each other
<jrtc27>
what doesn't kill you makes you stronger and all that
<xentrac>
no it didn't, it led to the worst period of colonialism in human history, with atrocities that dwarfed the holocaust
<xentrac>
in the same spirit, I think the current ransomware pandemic will likely do more to advance computer security than all the research we've done over the last 40 years
<xentrac>
I hope civilization somehow survives
snowrichard has left #riscv [#riscv]
Jmabsd has joined #riscv
<GenTooMan>
it will also likely get a lot of ransom ware creators physically dead.
<xentrac>
(it was Maxim's version of the machine gun that was really decisive for the colonial era, I think)
<xentrac>
oh, they'll all die
<xentrac>
the humans do that
snowrichard has joined #riscv
<xentrac>
the question is what kind of world they'll leave behind
<xentrac>
hopefully not one where every human on earth is constantly monitored through their cellphone for behavioral anomalies that could indicate their political loyalty is faltering, or something like that
<GenTooMan>
well email spamming took a dark turn when a prominent (arrogant) spammer in Russia ended up dead beaten to death in his home, likely because he wasn't "sharing" his wealth. Since then spammers have started to be a lot more mindful of what they do and the consequences.
<xentrac>
they haven't, no
<xentrac>
last spam I got was 5 minutes ago
<GenTooMan>
:D that's now then it dropped 98% instantly it took 2 years before it was even close to what it had been.
mahmutov has quit [Ping timeout: 252 seconds]
snowrichard has quit [Quit: Lost terminal]
<GenTooMan>
that said it would be nice if people tried to be mindful of other people.
<xentrac>
it was dark from the beginning. Canter was disbarred for illegal advertising practices 3 years later, Siegel died at 52
<xentrac>
but I'll believe the spam problem is solved when companies all go back to hosting their own email
<xentrac>
oh well. enough Old Man Yelling At Cloud
<xentrac>
these are going to be very interesting years, and I'm sure glad I'm not in any way associated with computer security
<xentrac>
well, except for the current DARPA project, which is hopefully minor enough to not matter
___nick___ has joined #riscv
<GenTooMan>
well each person who thinks they are in control are only fooling themselves.
<GenTooMan>
eclipse seems to be going through every bit of code i've written it still has 3500+ files to go ... what the heck.
<GenTooMan>
or is that projects.. it's been a while.
<pierce>
<xentrac> "but I'll believe the spam..." <- I'd love to hear more about this, I don't see how personally
snowrichard has joined #riscv
snowrichard has left #riscv [#riscv]
snowrichard2 has joined #riscv
vagrantc has quit [Quit: leaving]
snowrichard2 has quit [Client Quit]
devcpu has quit [Quit: leaving]
<xentrac>
pierce: hm?
<xentrac>
before spam it was normal for every company to host their own email. in between filtering out incoming spam with an acceptable false-positive rate, and not getting their outgoing mail marked as spam by increasingly aggressive spam filters, that is very much a minority choice now
SpaceCoaster has joined #riscv
SpaceCoaster has quit [Remote host closed the connection]
freakazoid343 has joined #riscv
drmpeg has left #riscv [#riscv]
freakazoid333 has quit [Ping timeout: 250 seconds]
drmpeg has joined #riscv
PyroPeter has quit [Ping timeout: 246 seconds]
PyroPeter has joined #riscv
geranim0 has quit [Ping timeout: 252 seconds]
<pierce>
Spammers could easily circumvent this by spinning up their own mail servers I'm sure
<pierce>
(not a security expert though)
<xentrac>
well, that's exactly why newly spun up mail servers are at a high risk of having mail from them rejected
<dh`>
even ones that aren't new
<dh`>
as far as I can tell gmail's default behavior is that anything from outside it is spam
devcpu has joined #riscv
jimbzy has quit [Ping timeout: 260 seconds]
riff-IRC has quit [Remote host closed the connection]
riff-IRC has joined #riscv
[itchyjunk] has joined #riscv
[itchyjunk] has quit [Remote host closed the connection]
jimbzy has joined #riscv
jimbzy has quit [Changing host]
jimbzy has joined #riscv
peepsalot has quit [Read error: Connection reset by peer]