MINTIA (not vibecoded)

[cursive charm]

this was apparently the source for this quote also from said friend:

Although, somewhat ironically, Microsoft allowed IBM to develop the PowerPC builds of Windows NT, which was where hibernation was first implemented in Windows

the PPC port was done by MS, so i didn't understand what they meant until they linked that

[twilit smelt]

sharing the same partition ID is probably some kind of result of HPFS being older than NTFS

often the way you start writing something like a filesystem driver or filesystem format utility is to copypaste another one and then remove everything but the skeletal commonly applicable bits

which in that case could have included the partition ID constant or something

just overlooked

[cursive charm]

that makes sense

speaking of utilities, i wonder if much of the filesystem utilities could be made freely source-available

i mean, i guess sources to them are present in the leaks

for some reason, the copy of the design workbook from the WRK has the coding style standards for the NT utilities, but not the design overview of ulib itself

[twilit smelt]

i enjoy how in minoca (which appears to be a (successful!) attempt to make "NT but cleaner") the filesystem drivers and filesystem utilities share the same core code

and for the utilities theres like a mock IO system for it to interface with which redirects the requests to a [device] file

[cursive charm]

oh, that's neat

i guess, in NT, all of the filesystem utility code is written in C++ and the user-mode fs implementations are dynamically loaded from a DLL (e.g., untfs.dll) for things like chkdsk

idk how it worked in old UNIX, but modern unix-likes would've had a separate program for each filesystem

actually, wait, i just realized: didn't they write parts of the configuration manager, specifically loading and parsing hives, to be usable in user-mode?

like offreg.dll in win11 25H2 has some symbols that match ones found in the kernel (e.g., HvpGetCellPaged)

so i'd imagine they did something like that but for loading hives, i guess

[cursive charm]

welp, i guess my bias towards casey didn't stop me from getting confused over a recent post of his: https://fixupx.com/cmuratori/status/2050496898889630164

Casey Muratori (@cmuratori)

At this point I feel like I should do a stream tomorrow to talk about the replies I've seen to this post. I completely disagree with people's umbrage about use of FPS as a metric here: A) that is exactly what time-to-show actually is (we measure 1% and .1% lows in for a reason!), and b) to me, FPS is the most relatable number for response time for average people to understand given that they don't work on software performance for a living like I do.
︀︀
︀︀Many people (especially gamers!) intuitively know what 10 or 11fps responsiveness feels like for an action. Few intuitively know what "94ms" responsiveness feels like.
︀︀
︀︀I also find it unacceptable to call this "load time" because the user is not asking to "load" anything - it is an action they are taking from a UI that they perceive to be contiguous, and the choice to involve a "load" of any kind at this point is purely the fault of the designers of the system, not some inevitability. Everything ha…

FixupX · May 2, 2026, 8:45 AM

in case you're wondering, here's his original post: https://fixupx.com/cmuratori/status/2050328300745261395

Casey Muratori (@cmuratori)

Just want to make sure I'm reading this right: Microsoft rewrote the run dialog with performance "top-of-mind", and the best they could manage to do when putting up a single text box was 10fps?

**💬 132 🔁 65 ❤️ 1.7K 👁️ 128.4K **

FixupX · May 1, 2026, 9:35 PM

yes, apparently, he uses FPS as a measure to describe the time it takes for the new run dialog box to show

hurray?

[earnest zenith]

I wouldn't take Casey Muratori seriously at all

his points are often bogus, his solutions inpractical, and he fails to understand the need for complexity in many cases

[cursive charm]

i honestly expected a little better from him than to see him double down

like sure, the new run dialog takes 93 milliseconds too long to show, but to describe it in FPS?

[earnest zenith]

the intention was probably to compare it to a game rendering a dialog box

and give the impression that all it does in those 93ms is render the dialog box

when in reality, there's other stuff going on

[cursive charm]

okay, well, he isn't talking about rendering anything. this is just about the time it takes for that run dialog to load and appear.

[earnest zenith]

yup

[cursive charm]

and in this case, it is actually "loading" something because the shell spawns a shellhost.exe instance to host it

casey could complain about that at least, but instead, he's like, "10 FPS was the most they could do?"

even then, he doesn't have to dumb it down to that degree

here he is being more clear: https://fixupx.com/cmuratori/status/2050330091386020277

Casey Muratori (@cmuratori)

↩ (@Shinobu_uwu)
@Shinobu_uwu It should be instantaneous. If the monitor was at 120hz, then that times the current compositor buffering is how long it should take to open a dialog. That'd be something like 16ms here, so assuming this number is accurate, they're ~5x slower than what I would expect.

**💬 2 ❤️ 92 👁️ 6.1K **

FixupX · May 1, 2026, 9:42 PM

i don't wanna know how he'd describe the time it takes for visual studio to start to a non-programmer

[dense vigil]

To be clear 93ms to start that window is absolutely ridiculous anyway

[cursive charm]

it was apparently an improvement from the previous run dialog, which was like ~110ms, according to sylv?

either way, i agree it's ridiculous

i just don't get the insistence on using FPS to describe it

[dense vigil]

Yeah that was pretty dumb

[cursive charm]

apparently, he compares milliseconds to a monitor's refresh rate: https://fixupx.com/cmuratori/status/2050509112853864515

Casey Muratori (@cmuratori)

↩ (@takeshi_no_uta)
I'm not saying it is fine or not fine - I'm saying that if they are claiming they rewrote the program for performance (the claim was that perf was "top-of-mind" during the rewrite, as I quoted), then 94ms is not a very good outcome - especially given that the previous version was basically the same (about an 8% difference, apparently).
︀︀
︀︀As for milliseconds vs. FPS, this is a post I made. I expressed the action perception the way I normally do when talking in public about the feel of interactive things. I find that FPS is easier for people to understand than milliseconds - and in fact I think of it that way in my head often, too, as a reference point (33ms is like a 30hz monitor, 16ms is like a 60hz monitor, etc.).
︀︀
︀︀The 94ms number was literally right there in the post as well for anyone who wanted to see it, as was the phrase "time to show". So literally anyone who wanted any version of any of these things could have gott…

FixupX · May 2, 2026, 9:34 AM

[shadow ridge]

1/10th of a second still makes more sense then 11fps

fps only makes sense when something is continuous

a dialogue box opening is not a continuous event

[short owl]

As if i needed to open the run dialog multiple times a second 😭

[twilit smelt]

i dont know it depends

operating systems arent games

games try to optimize for performance for themselves, operating systems try to optimize for performance for everyone else

i dont think casey understands this

93ms sounds right to me if its like loading a whole program for it or something

like a cold start

[dense vigil]

You of all people probably know what amount of work a computer can do in 93ms

[twilit smelt]

is it doing IO? like is it not in cache?

[dense vigil]

Its not a cold start, its a median they got

[twilit smelt]

i see

[dense vigil]

Its a damn window with an input box lol

[twilit smelt]

it really depends exactly how they implement it

[dense vigil]

I mean yeah obviously they aren't making it up, im just extremely curious what its doing under the hood that takes so much time

[twilit smelt]

first of all it doesnt really matter because nobody cares how fast their run box pops up

so like

i agree there could be vast wasted cycles from like

layers of OOP lasagna or something

but also

that might simply be the best way to do it

since performance doesnt matter

[dense vigil]

Well they were the ones who started bragging about it being remade with perf as top priority or whatever

[twilit smelt]

so then youd optimize for maintainability and whatever

instead

well thats a weird thing for them to prioritize

for the run box

[dense vigil]

They pointed it out because people were shitting on windows being an electron app lately

[twilit smelt]

and also making it take up as little resources as possible while its not in use

which means when it becomes in use you have much more set-up work

than like popping up a box in an already loaded game where they make sure everything possible is in memory at once and whatever

i think casey is looking at windows and applying his gamedev brain to it and that just doesnt work

[dense vigil]

The real reason is its probably 10 layers of frameworks and XML parsers etc

[twilit smelt]

yeah which would (or could) be terrible if it were a game

but its not a game

i think he should stfu and stay in his lane prob

[dense vigil]

Yeah but like, dont brag about it 😭

[twilit smelt]

yeah thats weird

but him measuring it in FPS is the most meaningless dumb shit ever

[dense vigil]

Yeah idk why he did that

[twilit smelt]

if performance were important i guarantee you they could get the run box to pop up 100s of times per second

this is obviously a "its because it doesnt matter" thing rather than a "its because theyre incompetent as opposed to him who is so smart" thing

he did this with the windows terminal too

he looked at theirs and it could "only" print like 100k characters per second or something

and so he like summoned an angry pitchfork internet mob of gamers to attack the windows terminal team

and made his own that can print 5 billion characters per second

or whatever

and that was supposed to be proof theyre incompetent

[dense vigil]

The windows terminal team told him fast character printing was an unsolved problem that required a phd

Which was absolutely ridiculous

[twilit smelt]

he aggressively misread that on purpose

[dense vigil]

He started by just making a GitHub issue

[twilit smelt]

a very sassy annoying github issue

to which they replied "this doesnt matter, we're not going to use phd code to make it faster"

[dense vigil]

So he did prove them wrong with that right

[twilit smelt]

the github issue was an attention stunt to begin with

and they could tell

they sassed him back slightly and he took that and ran with it

[dense vigil]

How do u know such details lol

[twilit smelt]

because i was watching it while it was happeninga nd this was all obvious

[dense vigil]

Eh idk

[twilit smelt]

this wasnt a rigorous engineering statement being made by that guy

it was an off-the-dome retort to some rando showing up and calling them stupid for having a plenty fast terminal

blowing it up like that wasnt fair or good faith and was done out of attention seeking and narcissism

it pissed me off bc the windows terminal team was doing objectively good and useful work and he had to come and swing his dick around over shit that didnt matter and publicly shame them to an army of retarded gamers

its not always the good guy move to point out to a million ignoramuses that some people doing otherwise useful work are being passively stupid about a particular thing that doesnt matter at all

in fact the only motivation for doing that is narcissism

[raven drift]

100ms to launch a gui app sounds entirely reasonable tbh

like, sure, you could integrate it into the shell

but that has little advantage

because nobody will care

[shrewd jasper]

yeah ig his entire argument crumbles upon the question "how many times a day ru opening the run dialog to the point where it taking 93ms to appear is a problem for you"

[dense vigil]

His argument came from them bragging about that speed

As if its something crazy or fast

[shrewd jasper]

its microsoft talking sbout windows 11, ig u cant expect much when they're talking about speed, like yeah they bragged about it, but its like bragging about the speed of your (non electric) scooter, you shouldnt expect much, but I do see how that would be confusing for most non tech people

[dense vigil]

https://x.com/i/status/2050642457801716066 new tweet

Casey Muratori (@cmuratori)

I wanted to test this now-promulgated claim that "94ms time-to-show" - assuming that's accurate - would constitute Microsoft having "managed to improve the run prompt latency" that "nobody has had an issue with" in previous versions.

On a 60fps* HDMI frame capture from my

X · May 2, 2026, 6:23 PM

[shrewd jasper]

did he disable animations for that though?

cuz sometimes windows are rendered but still arent ready cuz the opening animation is still playing

[raven drift]

theres obviously a lot going on with starting a modern gui program

[short owl]

Aren't most of the native GUI things in Windows 11 made with C#

[raven drift]

loading the theme, loading fonts, parsing them, loading i18n, layouting that respects output device scaling etc

[shrewd jasper]

he's probably thinking like it's a game, as hyenasky pointed out

[twilit smelt]

the fact he only listens to his audience and other influencers has stripped him of his ability to reason critically

[acoustic sparrow]

that doesnt really take 100ms though?

unless you literally load every font

which you should not do

[raven drift]

i'm sure you can do it faster if you overengineer it

[acoustic sparrow]

i dont think you need to overengineer it at all

[twilit smelt]

woah

xnu still uses a global hash table for (object,page) lookup?

it uses a BIG global hash table for (object,page) lookup!

each bucket is 12 bytes

so if you have 64gb ram

youll use 201326592 bytes on page buckets

192mb

thats kind of a lot

[acoustic sparrow]

not rly

[earnest zenith]

to me Casey Muratori is a clout-seeking clown who thinks everything is a game

[raven drift]

what is this hash table for?

[twilit smelt]

looking up virtual pages within vm objects

[raven drift]

so it exists once per vm object?

[twilit smelt]

no

its global

its keyed by an (object, page offset within object) pair

[raven drift]

ah

[twilit smelt]

and they size it to be whatever the next power of two number of buckets is from the number of physical pages in the system

[acoustic sparrow]

to virt addr?

er, phys addr?

ok if xnu does it

[earnest zenith]

ofc one can criticize Microsoft for publisizing the 93ms thing in a vacuum

[acoustic sparrow]

guess im writing a hashtable

[earnest zenith]

without stating what exactly is being loaded from disk

[twilit smelt]

they might be the last people to do it

[acoustic sparrow]

because its so much easier than a radix tree

[acoustic sparrow]

ok but a radixtree is less fun

[raven drift]

picking the least power of 2 above #physical pages sounds dangerous if it's an almost exact match

[twilit smelt]

using a global hash table for this was common in the 80s/90s

[raven drift]

if you hit a load factor of 0.99 or so the perf will be bad

i guess it's alleviated by the fact that not all pages are in vm objects

[twilit smelt]

i was going to do per-object direct and indirect and doubly indirect pages similar to unix style filesystems

[raven drift]

but still

[raven drift]

that's like a radix tree with optimizations for small objects

[twilit smelt]

yes

making it per-object can cause lifetime issues though

like

with a global hash table

its easier to destroy the object before you destroy the pages

[acoustic sparrow]

yeah even good modern hashtables only work well at like 12/14

[twilit smelt]

xnu's vm seems kind of like a dinosaur in some ways

it might be because it has run on a much more limited range of hardware than the others so it hasnt had as much evolutionary pressure

the total lack of numa awareness seems really out of place for current year for example

but then, apple has no numa devices

other than the multi-die apple silicon chips which i assume have extra latency for cross-die unified memory access... but maybe it doesnt? or maybe if it does its so minor its not worth being aware of

hm im wrong

thats actually not numa either

[dense vigil]

You mean the vmm overall?

[acoustic sparrow]

they do the meme where they stripe cache lines between numa nodes iirc

[twilit smelt]

yeah

the global hash table thing is like especially antiquated

im shocked to see that in a modern kernel i thought that died out in the 90s

[dense vigil]

What's the hash table for?

[twilit smelt]

.

[twilit smelt]

.

[dense vigil]

Whats a virtual page?

[twilit smelt]

like for example a page from a file

[dense vigil]

So this stuff is for page cache?

[twilit smelt]

its for any virtual memory object

which includes the file page cache

but also shared memory

and also (in mach-style vmms) private memory

[dense vigil]

Ah yeah pitust said radix tree

[dense vigil]

So like how ext blocks work

Depending on the number of blocks in a file it increases the number of indirections to find a block from offset

[twilit smelt]

yeah

thats actually much older than ext

[dense vigil]

Probably, thats just the only algo i know where thats utilized

[warm mural]

uvm also does this

[twilit smelt]

another archaic thing about the xnu vmm is the fact it still has shadow object chains it needs to collapse

a thing that was solved in the late 90s

[warm mural]

so it's still basically mach vmm

[twilit smelt]

yeah

plus memory compression which is good

[cursive charm]

fwiw, casey claims that there were more bad blood that followed than just that "PhD thesis" comment: https://xcancel.com/cmuratori/status/1522471966929653761

[twilit smelt]

i fail to see anything wrong with that behavior either

[cursive charm]

tbf, i don't really know what he meant by "talk about wt" exactly

[twilit smelt]

crybully whining after victim does a little trolling in his discord

[cursive charm]

i guess they just went and brought up wt a few times?

it's been a while since i've read the original thread, but i never understood why he has such a problem with people acting dismissive towards him when dhowett and others said what they said because they thought he was the one being rude

he tries to walk around it by saying this on twitter: https://xcancel.com/cmuratori/status/1522671172688646144

which, to a detractor, essentially means, "i don't know how to work with others"

because apparently he needs to depend on people to be able to hash things out rather than be the mediator himself

not that i'm defending the "PhD thesis" statement thing or anything. i'm just saying that it sometimes feels like casey lacks enough self-awareness to notice when he's himself being pompous or dismissive.

[twilit smelt]

this whining message in particular is really cringe

hes like so mad that this guy hes putting on blast to an audience of thousands joined his discord

like thats the crime lol

[cursive charm]

also, this is irrelevant, but now i'm somewhat cynically curious as to how cutler would treat casey if he was on his team during development of NT 3.1:

[twilit smelt]

i also saw that at one point

ngl he'd probably work fine on that team aside from it being a huge culture shock to work on a kernel team

i dont think he's ever fully understood how osdev differs from gamedev

he thinks everything is gamedev

[cursive charm]

he's never worked on a driver, so i doubt he'd even offer much unless he was on the graphics team, i guess

[twilit smelt]

a delusion powered by his audience

who validate everything he says

[cursive charm]

well, wait, wasn't the graphics team trying C++?

i guess so. page 77 of showstopper! has gates convince whitmer to use C++:

apparently, this ended up pushing back the graphics team's work as a result

[warm mural]

that doesnt matter

im sure there were a bunch on the team (excluding the dec guys) who'd never worked on OS stuff

i think casey wouldve been smart enough to figure it out

[cursive charm]

well, he said "kernel team," not any other part of the group working on the OS

also, maybe that was the case for DEC MICA, but idk if that's true for NT

could be wrong?

[warm mural]

wdym

[cursive charm]

if we're talking exclusively kernel team, that is

[cursive charm]

you said there could be plenty of people on the kernel team who never worked on an OS beforehand

[warm mural]

yes

[twilit smelt]

i dont think thats true

there werent that many kernel guys and they all had backgrounds in kernel stuff afaik

especially in 1988

which was the dawn of the team

theyd all done kernel work on at least mica and vaxeln and some had done vms and/or unix

[cursive charm]

i think cutler said in plummer's interview that there were several people on MICA who had no prior experience, but that's a far cry from whether there were any such people on the NT kernel team then

[twilit smelt]

i think all the ppl from mica who went to microsoft had also done vaxeln

https://bitsavers.org/pdf/dec/vax/vaxeln/2.0/VAXELN_Technical_Overview_1986.pdf

it sounds like the vaxeln team basically smoothly transitioned into being the mica team

modulo a few people

[cursive charm]

casey worked at microsoft as an intern around 1993 and/or 1994, so i guess he would've worked on some library for .ani files according to wikipedia

[twilit smelt]

btw, i always heard that vaxeln was where the kernel vs executive layering was first conceptualized that ended up in NT

it also seems to be where the habit to describe Ke's size in bytes of compiled code, that persisted through to the end of NT 3.1's development, came from lol

i immediately fell in love with NT's history back in like 2020 when i first read the workbook and saw in ke.doc that "We anticipate that Ke will be no more than 30 kilobytes of compiled i860 code." or whatever was like a major thing they decided to dedicate a sentence to

very quaint and indicative of where they were coming from and the crossroads in like 1989 when that was written in there

[cursive charm]

i recall one of the whitepapers microsoft publicly released regarding NT being a summary of ntdesrtl.doc from the design workbook

the original had an overview of branches of UNIX at the time and fragmentation involving APIs on various UNIXes

of course, it followed with this:

Microsoft cannot afford to present POSIX and OS/2 integration as poorly as most of the UNIX vendors have.

and then i guess they gave up on POSIX sometime after the switch to win32

[cursive charm]

i'm guessing that'd be this sentence from the end of the first section of the WRK copy:

It is estimated that the kernel will be less than 48k bytes of resident nonpageable code exclusive of the IEEE exception handling code.

i wonder what the original draft would've said

smithsonian hasn't really released the contents of their copy of the workbook

[twilit smelt]

yeah this is sad

their copy is dave cutler's copy btw

his OG copy

[cursive charm]

yeah

i remember you were able to get CHM to scan the DEC MICA design documents before, right?

[twilit smelt]

yes and also the ozix ones

they are in the process of digitizing their entire collection though

so really all i did was

move them to the top of the priority list (behind everyone else who asked for a specific thing before me)

@warm pine also i realized we actually do have a source on when ozix was cancelled

assuming thats based on a real thing that happened and not just him being annoying

it was cancelled almost immediately after it was designed

nvm

the design docs we have are from oct 1989 not oct 1988

so he was just being annoying

very presciently

because it did get unceremoniously cancelled

at some point

oh wow

the chm seems to have digitized a TON of DEC records quite recently

[twilit smelt]

https://www.computerhistory.org/collections/catalog/600000015/record/102781227/

thats interesting

rob short

decwest

april 1988

this is gonna be about prism and mica

found an ozix reference

WOAH

the CHM's new catalogue has the ability to search within all their pdfs

and im finding all kinds of stuff

[cursive charm]

i'm kinda curious to learn about the microsoft-DEC partnership thing

there isn't really much information on how it came about, afaik

that is, unless i'm just missing a lot, i guess

[twilit smelt]

wait

could this be the answer to where the mysterious "AXP" came from

VAX-P

V-AXP

[mortal thunder]

no they didnt?

posix subsystem had been a thing until (and including) NT 4.0

and i'm pretty sure NT still features support for duplicating/forking processes to this day

[cursive charm]

i didn't mean they cut POSIX out entirely. i meant that they scaled back their plans on POSIX sometime during development. originally, they had once considered shipping a version of NT with just the POSIX bits, going by elsewhere in the workbook, iirc.

and yes, process cloning still exists in NT, though mostly used for reflection nowadays, afaik

[cursive charm]

@twilit smelt, do you remember which part of the design workbook said this? i assumed it was mentioned in basecont.doc, but couldn't find an exact reference there.

[twilit smelt]

It's quite possible that's something I inferred from multiple disjointed sources like possibly the workbook plus antitrust case docs plus maybe emails with markl

[twilit smelt]

i love these internal assessments of the competition

why is pa-risc so dismal

2

[tame phoenix]

yes, but both assumptions are wrong, because the run dialog isn't the kernel or a game, it's a graphical shell

for a graphical shell the user expects responsiveness BUT its animated, so you have to factor in the length of animation

if the animation takes 100ms measuring the response time in FPS makes no sense

[twilit smelt]

did you think i was saying the run dialog is in the kernel?

[acoustic sparrow]

well it should be in explorer.exe right?

or the start menu

[tame phoenix]

no you implied it's part of the os, people here would debate that

[acoustic sparrow]

and then its instantish

[twilit smelt]

the run dialog is absolutely part of the os

[tame phoenix]

I think it is others say os dev is only kernel

[acoustic sparrow]

tbh i think 90ms is a weirdly long time for this anyway

the only thing that would take that long is a really slow font load

but like the solution there is to make font loading faster

[tame phoenix]

it's only weirdly long if it's noticeable to the end user...

[tough gate]

hello

[tame phoenix]

I bet the end user can't tell a less than 0.1s delay on a popup box

[acoustic sparrow]

this is true

but i do wonder why

[acoustic sparrow]

assuming you have a fast system, anyway

[tame phoenix]

because it has to render it?

[acoustic sparrow]

rendering doesnt take 100ms lmao

[tame phoenix]

does windows still render window decorations in kernel?

wouldn't surprise me if run dialog is a separate program these days though that it has to launch, you run a program to run a program

[acoustic sparrow]

ok but thats like still not that expensive

hold on im gonna get obs so i can compare to linux tools

[tame phoenix]

depends, it could be a .net assembly even

[tough gate]

why are you guys talking here

😭

[tame phoenix]

why not

[acoustic sparrow]

that would actually make it start slowly

and that might actually be why

that its .net or react native or something like that

[coarse current]

this is a text channel so people use it to talk

[acoustic sparrow]

i think thats a more convincing reason

[tame phoenix]

they went full on .net obsessed around vista and walked a lot of that back due to performance issues

[acoustic sparrow]

lololol

[tame phoenix]

that's why longhorn had problems

so I heard

that and project mismanagement

[shadow ridge]

this channel should be a 2000s IRC chatroom

just #mintia

[tame phoenix]

it's .net with winui 3, xaml

apparently

rendering it requires parsing a markup made of xml

that's what I could find but also you would have to load the command history from disk, to populate the drop-down

[acoustic sparrow]

yeeah

[acoustic sparrow]

well no you can do that in parallel

but thats fast anyway

fuzzel is ~1-2 frames at 30fps on my laptop

[acoustic sparrow]

the framework is 100% the reason for the time tbh

dotnet isnt terribly fast at startup iirc

[tame phoenix]

that's what 66ms?

2 frames at 30fps

[dense vigil]

I want to remind you that the 93ms is the median with hot cache

[acoustic sparrow]

yes but its dotnet runup

[dense vigil]

So you have to restart thousands of times to make it fair

[acoustic sparrow]

i ran fuzzel twice on hot cache

[tame phoenix]

hot cache doesn't mean it doesn't need to re-parse the XML though

[acoustic sparrow]

im not running it thousands of times thats like impossible to test

[tame phoenix]

it just means it doesn't need to load it

[dense vigil]

Yes, that means disk io is eliminated

[tame phoenix]

XML is icky smelly

[acoustic sparrow]

yeah but its probably not the cause either

xml parsing is fast

[tame phoenix]

fast compared to carrier pigeon

it's still a load of string ops

[acoustic sparrow]

nah its like actually fast

[shadow ridge]

isnt basic text parsing basically O(N)

even symbol table stuff only require like string hashing

[dense vigil]

I mean idk about XML, but there are json parsers capable of parsing many gigabytes if json per second

[tough gate]

plz stop talking here 😭

[teal trench]

do you come with fries

[acoustic sparrow]

no u

[acoustic sparrow]

i think xml is similar

[warm mural]

they are doing SIMD json parsers now

its crazy

[acoustic sparrow]

simd xml is also a thing apparantly

[warm mural]

isnt xml a pain to parse

[shadow ridge]

most of lexical analysis is just a while loop running a switch case

and the most complex op u ever do is like running isalpha / isdigit or whatever on the strings

thats probably not too hard to SIMD

[twilit smelt]

why are you so concerned

[tame phoenix]

a friend of mine is strongly competitive and set his mind on making something faster than simdjson

https://github.com/RealTimeChris/Jsonifier

comes with benchmarks but personally I'm still a big fan of nlohmann json. it's not fast but it's a very nice API.

[warm mural]

did he do it

because simdjson is worked on by top people

[tame phoenix]

his benchmarks said so

but i didn't use it

[tame phoenix]

https://tenor.com/view/top-men-raiders-top-men-raiders-of-the-lost-ark-top-men-indy-top-men-indiana-jones-gif-11904711898586575231

[balmy zealot]

Is this analogous to the Windows page frame database?

[twilit smelt]

no it is not

the pfndb is for physical pages, this is for virtual pages

[hybrid condor]

tbh I'd say XNU as a whole

[digital pivot]

its expected to be highly contended on lookups?

[twilit smelt]

https://raw.githubusercontent.com/xrarch/books/main/documents/xlospec/main.pdf

restyling the docs

[twilit smelt]

iokit isnt half bad

[dense vigil]

pretty

[hybrid condor]

iokit is pretty solid (it's not that the other parts are bad, but xnu might turn out to be more of a dinosaur and rougher around the edges than you thought, and it has more issues than you might imagine)

[hybrid condor]

actually page_lookup is pretty hot

in the zero fill path the release barrier can stall when there is an in flight vm_page_zero_fill

s

[twilit smelt]

i want to take a third shot at compilers

first one was old sdk which was the really really bad df compiler in lua

second one was newsdk which is the normal amount of bad (for a hobby compiler) in jackal for jackal

third one would be a better jackal compiler

maybe a general compiler infrastructure like llvm or like @sandwichman's iron

[mortal thunder]

its gonna take at least another year to finish mintia2 if you proceed

[twilit smelt]

i was going to prioritize mintia2

[shadow ridge]

do it 10 more times and you will become me

but you also wont have a mintia 2

[twilit smelt]

just found the part in the internal communications during the development of dec alpha where they reinvented the bottom-up page table walk

sitll didnt take it as far as i did though where it walks all the way up

[digital pivot]

how do you find all that stuff

[shadow ridge]

he requests companies to digitize magnetic tapes

then spends all his time reading it

while listening to the same song on loop

[digital pivot]

do they ask for money to digitize ?

[shadow ridge]

idk

[shadow ridge]

ok actually i dont think this is digitized tape

legit where do u find this

@twilit smelt

[twilit smelt]

OOH LA LA NOW WE TALKIN

[shadow ridge]

i dont think he even registers that we're here

he's in the internal doc grind

[digital pivot]

is it because apple core os is more busy with security rather than scalability or other matters of the core components like mm?

[hybrid condor]

Apple CoreOS doesn't do security

they wouldn't be able to differentiate between a security mitigation and an API boundary

[twilit smelt]

who up damaging they dec alpha 21064 by allowing a pin to rise above 4.0V before the 3.3V power to the chip is stable

[twilit smelt]

im listening to the boards of canada societas x tape

https://www.rareboc.org/Societas x Tape [2022-03-13].mp3

[warm mural]

Wait until mintia gets better and write the new compiler on it

ultimate larp

[twilit smelt]

i like how dec documentation always lapses into ascii art for bit fields

[twilit smelt]

the current compiler is already fully self hosting and runs on fox32os

[warm mural]

Yes but that's not what I'm saying

I'm saying you should focus on mintia then write the new compiler on mintia

[twilit smelt]

its so crazy theres like a whole other person also making a fake computer and that i wrote a compiler that runs on their unrelated fake computer as well as mine

[twilit smelt]

developing mintia2 on itself is the dream

id probably only do that on amd64 mintia2 though

[twilit smelt]

somewhat esotericmaxxing to listen to something i can only find as an mp3 on a random website while reading thorugh ancient internal corporate docs from 1990

[warm mural]

Performative smh

[twilit smelt]

these are my favorite songs on that mixtape

or most fitting for this activity at least

12:10 and following one
31:44
57:04
1:06:42
1:37:18
1:51:45
1:55:24

almost all of them are unreleased tracks shazam doesnt know

so i can only give them as timestamps

into this mp3 file

ur welcome

[keen stone]

is there anything more niche than this

[shadow ridge]

atonal music

[twilit smelt]

this isnt niche at all this is a mixtape bya n extremely popular band

boards of canada is like taylor swift for autistic zoomers and 40 year olds who went to college

they have niche aesthetics though which makes them popular among larpers like myself

[shadow ridge]

you should build an xr station mockup using like old microwave parts

then take a photo

using like a very old camera

develop it in a red room

then digitize it

you could collab with a film student from the same uni to make a 90s style documentary about how mintia was developed

you arent larping hard enough

[twilit smelt]

https://www.youtube.com/watch?v=TcZ_hMGQ-W0

i would just use footage from this

in fact i already used that footage for a vaporwave edit

https://www.youtube.com/watch?v=HyyuwhBkY8k

[shadow ridge]

no you need to build a set with like

very old computers around

people sitting on shiny XR stations

Jade Butler leading the team

blackboard with like kernel design stuff written on it

[twilit smelt]

lol

https://fxtwitter.com/taste_of_tbone/status/2051372605169418313?s=20

timothy faust 🇵🇸 (@taste_of_tbone)

calling something "retarded" in your investment firm documents is the conservative version of doing a land acknowledgement in your nonprofit budget report

Quoting ☔🔥☔ (@kirbywinfield)
︀
crazy how calling stuff “retarded” made the leap from comedy podcasters to the a16z style guide
︀︀
︀︀but you don’t ever hear these brave and hilarious vcs calling stuff “gay”
︀︀
︀︀makes you wonder.

**💬 1 🔁 12 ❤️ 125 👁️ 3.9K **

FxTwitter · May 4, 2026, 6:45 PM

[twilit smelt]

found this in my inbox from 11 years ago

looks like ive been on this app for a while

[shadow ridge]

nerd

all the cool kids are on TS 3

imagine using discord

[twilit smelt]

that is an 11 year old token and i doubt it could be used for literally anything ever but im still going to delete that just in case

[twilit smelt]

i like the completely unformatted 2015 emails

[warm pine]

and thus you settled in to 11 years of happy gaming, courtesy of your friendly neighbourhood discord marketing dude

[twilit smelt]

https://www.youtube.com/watch?v=PUc49j2VpZY

[twilit smelt]

he worked there until 2023 and then retired early at the age of like 30

as a multi millionaire

also the CHM has a bunch of Ken Olsen (CEO of DEC)'s personal correspondence

like TONS of it

and i was going through it and was stunned that like

PRISM only comes up once and its him off handedly going like "Our RISC project, which I'm told is referred to as 'PRISM', is one of the options we are considering for our future hardware direction, along with its software package known, apparently, as 'MICA'."

and its from like 3 days before it was cancelled

good to remember that DEC was a gargantuan company and that PRISM wasnt the most important thing going on at the time

at least they didnt think so

whole thing happened under the CEO's nose, not as some kind of skunkworks but just because it didnt draw his attention until like 3 days before it got canned

lol

now in retrospect we know that PRISM was actually the most important thing goin on at the time and cancelling it was potentially fatal for the company

but Ken had no way of knowing that

he was still looking at things like VAX was their eternal cash cow and the RISC stuff was some kind of uninteresting hedge going on in the background just in case

he got rid of it at the suggestion of the first naysayer about it in his inner circle

which is the only reason he was talking about it at all

and said naysayer was immediately like, after prism was gone and cutler left, "What if we do a new architecture, that's a kind of Extended VAX, or EVAX?"

and ken is like "wow great idea!"

and then they did prism again immediately except years late and called Alpha

that guy kind of couped dave cutler as leader of the risc effort, didnt he

at the expense of the company and taking advantage of ken olsen's narrowsightedness

really feels that way

bob supnik

in his defense when he talks about it in his chm oral history interview he looks almost like hes trying to avoid crying

it was all very hard and dramatic

its possible he just made an honest mistake and then fixed it after it was too late rather than it being some kind of cutthroat political move

[twilit smelt]

https://youtu.be/T3tcCBHRIfU?t=9702

@warm pine very interesting context for the prism stuff^

2:41:42 if the timestamp doesnt work

[twilit smelt]

this is interesting

[twilit smelt]

@warm pine indirect ozix mention

"megasafe file system" - msfs, aka advfs

"built at decwest" home of ozix

[mortal thunder]

i only joined discord in 2017 (partly because i would've been too young earlier) but i remember when they included random fun facts about random video games:

[warm pine]

I wonder why they renamed it

Its a bit sad really that it was added silently with no mention of its heritage that im aware of

Speaking of Ozix I have just realised that the Netware architecture appears to resemble it, with "protection domains" and automatic gate crossing. I do not have an explanation of this yet

[twilit smelt]

it seems like throughout 1989 ozix was considered a part of the alpha project but by the end of 1990 it has disappeared

there is no mention of its cancellation which apparently occurred some time in 1990

or what happened to it

it just stops being mentioned aside from in newer versions of older documents where nobody has removed it yet

and then only in lists like "VMS, ULTRIX, and OZIX"

this is my impression from going thru like 1000 pages of alpha stuff in the last few days

they prob decided OZIX and OSF/1 were redundant

so they canned the former (being less developed) and rolled its best pieces into the latter

but there are no communications relating to that story anywhere in here

nobody who would have them has donated them to the CHM as yet

[warm pine]

the scale of it was so vast

MSFS alone is insane and prefigures ZFS in many ways, a decade earlier than it should've been

i will look forward to hopefully finding out a bit more about whether that sheer scale conditioned its cancellation

[twilit smelt]

do we have a list of names involved with ozix

we might be able to just email them

and ask

i think the obscurity here is just a function of nobody who was directly involved with ozix contributing any of their like archived memos or whatever to the CHM

so we only see ozix from the POV of like

the alpha hardware team

like the copy of ozix docs we have is copies that an alpha hardware guy happened to scrounge up

but he obviously wouldnt have personal communication or internal memos

for the ozix team

[warm pine]

https://who.paris.inria.fr/Gilles.Muller/papers/RR-2647.pdf

protection domains again!

this concept keeps coming up, i need to figure out if from DECwest or a common source

[raven drift]

PDs are mostly just another way of saying "processes"

[warm pine]

so it would be, if you take that they share an address space and selectively block out bits of it by actual page table manipulations, and you have shuttles/migrating thread style RPC with a common stack that appears in all of them

so i say the protection domain concept that was fashionable in that time and appears in OZIX, DP-MACH, and NetWare is halfway between a process and a kernel module

[twilit smelt]

https://www.computerhistory.org/collections/catalog/102717706/
https://www.computerhistory.org/collections/catalog/600000031/record/102781227/
https://www.computerhistory.org/collections/catalog/600000031/record/102781228/

3 tapes of presentations given by the PRISM group months before cancellation, currently inaccessible

according to a wikipedia citation the third one mentions ultrix running on prism

heres one i happen to have the link to in my browser history before it was rudely obscured from view from the chm's website

https://www.youtube.com/watch?v=ead4qggQDyM

thats a well dressed elderly woman going into detail on the current status of porting VMS to Alpha (despite the title)

development manager

VMS was always the girlboss OS

idk why there was just a spectacular number of women involved with its development compared to other OSes at the time and especially by the standards of today

[twilit smelt]

ohhh

its because of world war 2

hitler takes power -> there are a bunch of old ladies programming the VMS kernel in 1990

this is real

[shadow ridge]

i cant believe world war 2 ended in 1995

what a time

[twilit smelt]

no i wasnt joking

there was a huge influx of women in stem fields during ww2

these women got their start around then

[shadow ridge]

ok that makes sense

[warm mural]

Programmers were mostly women for a while because the job was thought of as the same level as that of a secretary

[twilit smelt]

this isnt that

os dev was never 'relegated to secretaries', it was always highly educated ppl from top uni math and physics departments

like this multics programmer from MIT

covered in source code listings

presumably debugging

thats how they would debug operating systems back then they would just stare at source code listings for like 8 hours

[warm mural]

I'm not saying they weren't educated

[twilit smelt]

im just saying these particular women werent there because it was considered a lowly job it was bc they were highly qualified

ur right that that was a thing in the broader space for a time in like the 50s and 60s

but in this particular sub area, operating systems, it wasnt

[twilit smelt]

currently reading a bunch of docs about OSF/1

to try to figure out why tf the PILLAR spec from 1989 opens with like

"This language was previously intended for the PRISM project and is now being used for OSF."

one unrelated thing at one point that i learned is that at one point there were talks of classifying Mach as munitions so that other countries couldnt get it becuase the US government was convinced it was like an overwhelming technological advantage

there seems to be a lot of talk in these documents of the "basis" for OSF being up in the air and not yet decided. for a time at least it was intended to be based upon IBM AIX. speculation: was mica/ozix at one point slated to be proposed as a contender for this?

interesting comparison between AIX and SVR4

[cursive charm]

i remember someone from ncommander's discord who has supposedly worked on OSF/1

[twilit smelt]

Was it mxshift

I got in an argument with that guy about whether NT was more like Mach or more like VMS and I was arguing that it was ultimately more like the former and I was using like primary source documents and source code arguments and he just kept name dropping ppl like "I could call Dave Cutler right now and he'd call you an idiot."

Which was annoying

I was right btw

If Dave Cutler said it's more like VMS he'd be wrong

he didnt even seem aware NT had direct mach influences and quietly conceded that part when i brought up receipts

he was bought into the pop history version where it was designed in a total vacuum as a direct follow-on to VMS or whatever

so there goes his claim to special knowledge

[cursive charm]

yes, it was mxshift

they seemingly left ncommander's server sometime ago for whatever reason

i would've asked them about their work on OSF/1 or something had they still been there

also, i thought they had a mach guy on the NT team, lol

[twilit smelt]

they did!

they even tried to hire Avie Tevanian for the role Cutler took

as head of the PSG working on Portable OS/2

he turned them down bc Steve Jobs gave him an offer for NeXT that was more appealing

also the first project the PSG ever did, which was pre-Cutler, was investigating Mach's viability as a new OS/2 kernel

which was a project called Psycho

that started in early 1987 i believe

they only abandoned this bc Cutler insisted on doing his own kernel with his team

[cursive charm]

"psycho" suddenly reminded me of "whitmer's undead," lol

[twilit smelt]

imagine if he said yes

tevanian says he turned them down for that role literally 2 weeks before cutler was hired to do it

cutler would have gone on to do his mach on mips workstation company that he was originally planning on doing, which likely would have gone belly up within a decade or so

Windows NT would have been mach

NeXT prob would have sucked more

[cursive charm]

i guess apple would've gotten beos or something

[short owl]

😭?

[warm pine]

it's official

mach is the bomb

[warm pine]

this one is interesting, C really ended up eating everything

my working hypothesis is something like OSF being a generic concept of an open unix and ozix was perhaps at one point considered? perhaps the OSF/1 we know was some sort of quick measure (note the /1) based on the already-functional mach and it ends up being the last survivor

[shadow ridge]

do you actually enjoy these arguments btw

like

ive never seen you ever actually change someone's opinion

and it surely cannot feel good to argue with someone and literally never arrive at a shared conclusion

[warm pine]

Its a compulsion

When someone is wrong about nt you've got to correct the record. To not do so feels awful

[night needle]

i have abandoned the idea that you can change someone's opinon of something by talking to them

[shadow ridge]

makes sense

[shadow ridge]

i think it's feasible depending on how much of an opinion vs fact it is

depends on the person too ig

also for some reason i have boards of canada in my youtube recommendations now

[twilit smelt]

Yeah from reading these late 80s to early 90s OSF correspondence and whatever that's absolutely what it was

They were considering multiple options

In fact it sounds like IBM's condition for joining was that they look into using AIX for it and then they were fully planning on doing so around 1989 but AIX's schedule kept slipping until IBM backed out and Mach became the #1 candidate

[twilit smelt]

I almost always do change peoples opinion I feel like

This is one topic people are highly changeable on

to the point where if 1 guy who should know better is stubbornly wrong I will remember and complain about it for 3 years

https://www.computerhistory.org/collections/catalog/102750562

@warm pine see here ^ for the early OSF stuff

ozix was not rly mentioned there although i did just find it elsewhere...

"Ultrix/OSF" and "OZIX/OSF"

in 1988 they called something "Firefox" and something else "ISIS"

what did they know...

[warm mural]

is firefox an english word?

or is it FireFox, as in "fox on fire"

[shadow ridge]

probably the latter

i dont think it's a word

[warm mural]

it'd be kinda crazy they came up with the same name if its not a real word

https://en.wikipedia.org/wiki/Firefox_(mythology)

[twilit smelt]

its the latter

[warm mural]

ah

[twilit smelt]

but like

in this case specifically

i think its because its something derivative of the internal DEC project Firefly

they called the follow-on Firefox

[warm mural]

ah

but it's still a weird coincidence imo, it's kind of a weird name

[warm mural]

maybe this is the source tho

and they (mozilla and the OS guys) just looked up fire in the dictionary

[shadow ridge]

waterdog

[warm mural]

well, GNU IceCat exists

[shadow ridge]

LOL

[warm mural]

it's a firefox fork

but gnu

[twilit smelt]

unrelated to anything but have any of you encountered the phenomenon where youll be having a normal chill multi-sided convo in a discord and suddenly someone will go "WHY IS EVERYONE FIGHTING!!!! PLEASE STOP!!!!!"

[shadow ridge]

YES

[twilit smelt]

and its always someone with a sonic pfp and like "14. I am autistic and aro ace." in their bio

[warm mural]

that's annoying

[twilit smelt]

or something along those lines

[shadow ridge]

i have experienced

literally this exact scenario

like multiple times

and it's always like a relatively chill conversation compared to how the server is most of the time

[twilit smelt]

its a weird thing i just dont have theory of mind for because it must require a completely different kind of autism from mine where someone will just randomly read a normal conversation as like extremely aggressive fighting and will like start crying irl and start begging everyone to stop

[shadow ridge]

https://giphy.com/gifs/tf2-team-fortress-2-pyro-Zljn1ldRjeFPy

[shadow ridge]

not this part tho

they usually do have weird bios tho

i will never understand people who put shit like "UwU cute femboy :3 ~" on their bios like

maybe theyre running undercover predator catching operations idk

[twilit smelt]

also more on this, one reason this is a very rewarding thing to be autistically fixated on is because its one of the few things you can actually usually change peoples minds on. because its not some divisive political issue its just like "ackshually thats not true. here is an extremely specific document from 1987 that directly contradicts you" and theyll go like "oh i didnt know that. where tf did you find that"

like people do actually change their minds 99% of the time once you present primary source evidence

[shadow ridge]

theyre right tho

where the fuck do you find these

like

at first i thought you were asking some company to redigitize magnetic tapes

but i dont think thats all

[warm pine]

at first i would assume the poor unfortunate is the son of a broken family

then i would reconsider and assume they are concern trolling

[twilit smelt]

the mxshift guy was annoying bc he would just ignore it and go "I had lunch with Mark Lucovsky more times than you've taken a piss you little brat." over and over and over

like it was very tied into his self image and ego that i be wrong and he be right

on that particular thing

due to his work history or whatever

[warm pine]

the frustrating element is perhaps that making the point of personal experience would seem quite reasonable in a lot of situations but it's being deployed not in the situation which calls for it

[twilit smelt]

"I had LUNCH with Mark Lucovsky!"
"Oh okay here's a document written by Mark Lucovsky where he says Mach is a huge influence: <link>"
"Again, I had LUNCH with Mark Lucovsky!"

it was just that on a loop

[shadow ridge]

mark lucovksy would feel disgusted at this if he stopped yelling at bots online

[twilit smelt]

not necessarily

[shadow ridge]

idk i would be pissed if someone's using my name to win an argument

[twilit smelt]

if he were looped in he might side with his buddy because thats more fun

[warm pine]

it could well be that even lucowski now thinks that VMS is more important and thus agrees with mxshift because human memory is fickle and changes over time

this is one of the problems

[twilit smelt]

its also very possible that im more interested in the history of those things than almost anybody who was there and so ive seen a broader range of primary source documents from a broader range of perspectives than any one of them ever had access to at the time

[shadow ridge]

idk like if i went on the keyronex thread and i was like "cuck chuck you are WRONG I had LUNCH with WILL LIMN" would you be comfortable with that

[warm mural]

id be surprised if will wasn't one of the people in the world with the most knowledge about OS history

[twilit smelt]

which could result in me having hot takes that would sound insane to a couple of them and which are also literally true

[warm pine]

there's probably very few people care to apply an historian's approach to the history of software to this sort of degree, and if it were done more often than it now is, then a lot of myths would be shattered

[twilit smelt]

like in the microsoft antitrust documents that the court subpoena'd (and later put on the internet for everyone to see) theres an email from bill gates to a high up intel guy in charge of i860 where he threatens to stop using the i860 for the NT OS/2 project if intel isnt sufficiently supportive and communicative

from late '89

and i sent that to markl and he was like "I'd never seen that before. I always thought it was just because of the technical deficits that we switched away from it"

[shadow ridge]

did u ever speak with dave cutler

does he even have socials

[twilit smelt]

no he doesnt talk to randos

[shadow ridge]

better than screaming at bots all day tbh

[warm mural]

dave cutler is also like 90 so I doubt he has any socials

[twilit smelt]

no that would be one of the people who was there AND has an autistic fixation on the history AND has ready access to everybody else who was there (due to connections from having been there) AND has been looking into it for decades

like the bitsavers guy

[shadow ridge]

he probably has an email

[warm mural]

well yeah

but that's not a social

[twilit smelt]

i probably know more than him about the specific topic of like late 80s DEC RISC operating system research but thats only because thats my special interest and he has other special interests

[twilit smelt]

i emailed him and i got an autoreply saying

"I am OOF until 1949. Things were really tough in 1949."

dont know what that meant

i emailed him again 2 years later and it changed to

"I am OOF until 2049. Things were really tough in 1949."

quite mysterious

[warm mural]

https://www.bitsavers.org/

[shadow ridge]

he was born on 1942

[warm mural]

almost 2tb holy shit

all scanned by hand by one guy 😭

autism final boss

[shadow ridge]

he would have been like 7 in 19489

[warm mural]

but that's very cool though

[twilit smelt]

its funny some of the records he has have

at the very top

"Al Kossow" as one of the people in the revision history

in particular some of the 90s Apple documents

thats him!

[warm mural]

whatd you email him about

[shadow ridge]

he probably sent him yaoi as usual

this is why markl doesnt reply to him anymore

[twilit smelt]

i think the first time i was just asking if i could ask him a couple questions and the second time i said something like "A lot of people are very interested in what you have to say and so you should do an interview or a blog like Raymond Chen's" or something like that

(as politely as i could)

[warm pine]

unfortunately a lot of them are on their way out now

[twilit smelt]

like 4 months later the davepl interview came out

so maybe i had an impact

[shadow ridge]

last living student of rachmaninoff died like a few weeks ago at 102

seymour bernstein also died like last week at 99

[twilit smelt]

those were years apart and having exhausted my 2 strikes i will not email the man again

ive met other ppl who also tried to email him and got the same autoreply they were also confused by

[warm pine]

i might have had the chance to speak to sian mountbatten (formerly of the royal signals and radar establishment, and of ICL, and would've known about OpenVME, an old mainframe OS i'm interested in) if i had known a few years ago, as she lived not 40 miles from where i grew up, but unfortunately she died a few years ago

[twilit smelt]

evidently he doesnt speak to the unwashed masses

[warm pine]

a lot of knowledge about computing history will be lost probably forever as the boomer generation dies off

[shadow ridge]

theres a chance he's demented and is like "NT? windows? idk about any of that young man. I need to go i need to go work on VMS (or whatever he worked on before NT im not a history nerd)"

[twilit smelt]

hes not

[shadow ridge]

he's like 84

how would you know

[twilit smelt]

he did an interview like 2 years ago where he was extremely sharp

and remembered things from 60 years ago vividly

i think hes fine

[shadow ridge]

oh

[twilit smelt]

https://www.youtube.com/watch?v=xi1Lq79mLeE

[shadow ridge]

he's STILL working?

[warm mural]

yeah and afaik he's still working

oops yea

[twilit smelt]

ive seen the bitsavers guy around discord a bit and hes one of those older guys who has zero cognitive defense from the internet due to not getting on it until he was like 40 so hes really sensitive

and will leave any discord hes in

on a dime

[warm mural]

well tbf he likes what he's doing and he's probably not getting bossed around a lot

[twilit smelt]

for the slightest annoyance

and never return

so whenever i see him somewhere im like ah someone is going to double ping him and he'll leave within 24 hours

hes like a rare bird

[shadow ridge]

man this reminds me that technically gen z are like fucking cockroaches

just unnatural resistance to ragebait

[twilit smelt]

we can only hope to work on operating systems for 60 years straight

[shadow ridge]

and incredible information filtering

[warm mural]

personally just anything which interests me

not necessarily OSes as long as its interesting and stimulating

[twilit smelt]

dave cutler went from like the magtape, paper teletype, core memory era

to xbox hypervisor

in his lifetime

working on everything inbetween

very unique career

[twilit smelt]

yeah its the same reason you cant really fault mark lucovsky for spending like 8 hours a day arguing with obvious bots on twitter

[warm mural]

and he started by working for dupont or something

[twilit smelt]

there are two outcomes for an older person exposed to real time chat on the internet and both of them involve being what we would consider really sensitive:

1. the bitsavers guy, leaving on a dime over the slightest annoyance with no warning
2. mark lucovsky, arguing nonstop for 8 hours a day with bots and any other ragebait that grabs his attention

its like a flip of the coin which one they end up being

davepl was doing #2 for a bit in his own discord server he was engaging with every single piece of obvious ragebait from random zoomers

[shadow ridge]

theres a comedy anime where an angel comes down to earth and instantly gets addicted to MMO RPGs and this reminds me of that

[twilit smelt]

until he suddenly stopped typing there at all

bob supnik was also doing #2 for the brief period he had a github account

people kept ragebaiting him in the issues of his projects

eventually he just deleted his gh account

its a sad thing about these folks is that they end up inaccessible to us bc theyre not innoculated to ragebait and so they end up having to seal themselves away

its not something they ever encountered before in their like half century of being alive is being in a room full of people constantly trying to annoy them into a reaction, who arent getting kicked out of the room or their lights punched out for it

[shadow ridge]

they're not as attached to the online space as we are so they havent built defense mechanisms to it

but they dont NEED the mechanisms

cuz they arent attached

[twilit smelt]

well sometimes they try to like reach out to crowds of ppl who would engage with them in good faith and they just cant filter the ragebait out from that

it requires brain circuitry they dont have

in a way im glad that davec does not ever engage with anyone outside of his like irl working sphere

because it would be too much to see him do a markl

[shadow ridge]

tbh the amount of brainrot and ragebait and trolling that exists today is recent like

i dont recall the internet being like this before the 2020s

[twilit smelt]

like this circuitry has never failed them in their entire life:

this person has insulted or annoyed me -> they are in the same room as me -> the room is related to my work -> therefore they must be relevant to me and my work -> therefore i must and should respond

[shadow ridge]

sure it still existed but it wasnt at this scale

[twilit smelt]

this is generally true in real life social situations, but not on the internet which is a whole new ballgame

so like if davec made a twitter he might go "Hey I'm Dave Cutler I wrote the windows kernel!" and someone might reply "Windows kernel freaking sucks!" and then he might feel like he has to defend his honor even though its a throwaway nothing reply, because he doesnt know intuitively that its a throwaway nothing reply and that everybody else who reads it who matters will read it as such and ignore it entirely, and that replying to it would only draw attention to it and legitimize it

and boom he's made a fool of himself in a way that he has never encountered before in his life

its a mode of interaction they just dont have training for so theyre like 80 year old middle schoolers

they cant be blamed at all of course they didnt ask to live through the invention of such a thing for the first time in human history

theres a movie where in the first half this happens to a like 40 year old guy who is a famous professional in his field and he becomes an internet lolcow

called Chef (2014)

the second half is him redeeming himself

[shadow ridge]

is it like a documentary movie

or fiction

[twilit smelt]

its a silly fiction movie

its a comedy drama

[shadow ridge]

makes sense

[shadow ridge]

this one is called Gabriel Dropout btw

[warm pine]

i think the bigger problem is the crowdsourced panopticon people are subject to nowadays

for a baby boomer having an embarrassing moment is still just the minor setback that they always took it to be, barring they are unusually neurotic

[twilit smelt]

zoomers may develop nerves of steel compared to previous generations due to the extreme consequences of embarrassing moments and the fact the #1 intra-zoomer activity is to ragebait one another in an attempt to generate said moments

much to learn from this

[tame phoenix]

25 what though, bacon per hour?

[twilit smelt]

https://fxtwitter.com/_chair/status/2052195196499116427?s=20

Chair (@_chair)

relatable

**💬 55 🔁 349 ❤️ 9.7K 👁️ 301.3K **

FxTwitter · May 7, 2026, 1:13 AM

you can text this in confidence to someone and then 5 years later have it leak and have hundreds of thousands of people laugh at it

[digital pivot]

once hyenasky asked something about some older cutler's os or whatever on another server, and some guy from that server who knows cutler irl asked cutler himself about it

[twilit smelt]

did that happen

i dont remember that

oh lmfao

this was the 'question'

not even really a question

[digital pivot]

probably seemed interesting for him himself

[twilit smelt]

very fun

[hollow musk]

I can just imagine how surprised he was

[cursive charm]

did they actually say that

i just remember them going "want me to call dave cutler" or something

[hollow musk]

if you don't mind me asking which server are dave cutler's friends on?

[cursive charm]

mxshift was on ncommander's discord server

[warm mural]

in the 90s NT team discord

jk that doesnt exist

I think...

[cursive charm]

i wish that existed

[twilit smelt]

ozix mention

didnt find this before because its misspelled as ozyx

march 1990

[warm mural]

plot twist: ozyx is the canonical spelling

[cursive charm]

okay, they apparently said they were friends with many of the "key personalities" involved, which is somehow a rebuttal, i guess

also, they actually go by they/them

i still share a server with them, it seems, so their bio loads on my end at least

[twilit smelt]

mica mention

(PVMS was mica's first name)

[cursive charm]

portable VMS?

[twilit smelt]

PRISM VMS

[cursive charm]

oh, i see

[warm mural]

mhm this mica thing reminds me of NT...

[twilit smelt]

its a coincidence im sure

[cursive charm]

i read structured condition exception handling

[twilit smelt]

its funny seeing memos from other people and their perspective on decwest and theyre always like "jesus fucking christ those decwest dweebs are trying to steal shit from us AGAIN"

#2 is wrong

[cursive charm]

interesting analogy for #2

[hollow musk]

sorry if this has already been discussed - I forgot...
I seem to recall you were aware that dave cutler is still working on the nt kernel and overseeing its lower layers?

[twilit smelt]

he is not

i believe hes doing hyper-v or azure or something

but still keeps an eye on NT

[hollow musk]

ah, what a pity

[mortal thunder]

maybe the alternate timeline where the better Windows was based on Mach wouldn't have turned out so different, after all, if NT ended up borrowing a lot from Mach anyway

[hollow musk]

I hope that's the case, and that he influences the development of the kernel even today; at any rate if I worked there I'd want a veteran like him to pass on some of his knowledge and experience

[digital pivot]

yeah i think in his interview with the task manager guy he talked about that

[twilit smelt]

@warm pine what if ozix was never actually cancelled and just kept getting pared back and eventually did "ship" in the form of a number of components of OSF/1

[warm mural]

<insert obligatory davepl is annoying message>

[warm pine]

not impossible i suppose

[twilit smelt]

ive been stunned at how i cant find any references to it being cancelled at all

it just stops being mentioned at some point

but that could explain it

a deep dive into OSF/1 to try to discover anything ozix-y (besides advfs) is probably in order

not the sources we have, those are too early

i might actually RE a later release

[warm pine]

interesting word pops up

"POLYCENTER Advanced File System"

[twilit smelt]

polycenter?

[warm pine]

mentioned several times to qualify advfs

e.g. in https://www.techmonitor.ai/hardware/dec_announcements_1?cf-view

[balmy zealot]

Bit of a random one, but circling back to something from earlier, back in Dave Cutler’s day, one person could apparently work on pretty substantial chunks of the kernel single handedly. As I understand it, Dave Cutler designed and wrote a large portion of the kernel and the hardware abstraction layer himself. Is that still realistically possible nowadays, or has everything become too complex and tangled for that?

[twilit smelt]

wha

i fucking hate how ozix only ever pops up as like just random little confusing mentions

nobody ever goes "This is what OZIX is. This is what it means to the company. This is why we're doing it..."

they just mention it like you should know what it is

of course we know what it is, we think, based on some documents written by the OZIX group themselves

but getting some other perspective would be really great

we need to just like reach out to one of these ppl

on linkedin or email or something

Roger Heinen would be a great person to ask

[warm mural]

one day you'll organize all this stuff and write an article and instant pulitzer prize im sure

[digital pivot]

i believe its still true, 1 person for the memory manager, 1 for the scheduler and so on

[digital pivot]

or a book.....

[warm mural]

even though he's in pretty deep, I have a feeling making a proper book would require way more info/insight, potentially by people involved

[twilit smelt]

from feb 1988 @warm pine

"SFS"

could this be a predecessor to MSFS

Safe File System?

also "modula-2+ influenced PILLAR"

[warm pine]

i wonder which bits specifically

PILLAR is bigger and generally more pleasant for systems programming than modula-3 which has the common disease that working with fixed representations is tedious (not impossible at all, but it's thunderingly obvious it wasn't designed by people strongly considering that use)

[warm mural]

I wonder if one day that family of languages will pick up again

like oberon

[warm pine]

i was sketching a compiler for my own answer to that question but doing compiler work when i wanted to do the OS got to be quite offputting in time

[warm mural]

what did you have in mind?

I'm thinking something like oberon-02 or whatever is the newest can probably be made more modern easily

[twilit smelt]

a manager guy recognizing a similarity between "the SRC research kernel" (Topaz) and P/VMS (Mica)

could this be a conceptual germ of OZIX...

from page number something of asset 2 of this https://www.computerhistory.org/collections/catalog/102750399

[warm pine]

an example of what code actually looked in what i was working on

PROCEDURE SignalObject(
  object: PTR DispatchHeader;
  VAR wakeQueue: WaitBlockTQ.Head
) =
BEGIN
  WHILE
    NOT WaitBlockTQ.IsEmpty(object^.waitblockQ) AND
    object^.signalState > 0i32
  DO
    VAR waitBlock := WaitBlockTQ.RemoveHead(
        object^.waitblockQ, FIELD_DESCRIPTOR(WaitBlock, tqEntry)
      );

    VAR r := TrySatisfyWaitBlock(waitBlock);
    IF r = TrySatisfyResult.SatisfiedWhileEnrolling THEN
      AcquireObject(object, waitBlock^.thread);
    ELSIF r = TrySatisfyResult.SatisfiedWhileWaiting THEN
      AcquireObject(object, waitBlock^.thread);
      WaitBlockTQ.InsertTail(
        wakeQueue, waitBlock, FIELD_DESCRIPTOR(WaitBlock, tqEntry)
      );
    END (* IF *);
  END;
END;

[warm mural]

mhm what about ^DispatchHeader for pointer types

[warm pine]

it was basically modula-3 - GC + fixed-size integer types + things i needed for generic intrusive datastructures + some pillaresque concepts (FIELD_DESCRIPTOR is one iirc)

[twilit smelt]

some compiler stuff

sprite mentioned

[warm mural]

holy shit that image is huge my bad 😭

[shadow ridge]

big sprite

[warm mural]

what's sprite though?

ah uc berkeley distributed OS

[twilit smelt]

ozix mentioned

(EVMS is just Alpha/VMS)

[twilit smelt]

1992

megasafe file system mentioned

[shadow ridge]

why is garbage collection repeated

[twilit smelt]

they forgot to garbage collect the second time

[shadow ridge]

will limnstation 2.0 have garbage collection

at cpu level

instead of a regular address space

you reference objects instead

so MOVs are more like offsets into object buffers

and CPU handles the allocation / release of memory

[warm mural]

Lisp machine vibes

[warm pine]

Limn might not have that but Linn did

linn being a well known scottish company dealing in sound hardware and hi-fis

they invented a new architecture called the REKURSIV which was just like that

https://en.wikipedia.org/wiki/Rekursiv

[twilit smelt]

i need to just accept that ozix wasnt that important

and barely existed as a blip

for like 8 months

and nobody talked about it

ever

and stop searching thru these thousands of pages

i need to like do work

[dense vigil]

but he made the task manager