<abathur>
julm yeah; I was a little disappointed with the clickbaity title once I actually read the Twitter thread and saw the buried lede that his objection was security claims
<julm>
abathur: clickbaity title indeed :/
<abathur>
which is also, I guess, trite in a security context
<abathur>
almost all of security is a weird trust-propagation shell game that has to start with trusting things vanishingly few people can prove are trustworthy
<abathur>
wHY bothEr ENcrYPTING yOur DATA If YOu can't proVE Your HaRDDRIve fIRmWaRe doeSn'T HAvE A bugdOOr?
<gchristensen>
lol
<gchristensen>
anyone know if readdir() returns a consistent-in-time dataset, or if I iterated very slowly, would it show me new files being created?
<gchristensen>
or some other strange behavior like that
* samueldr
is scared
<samueldr>
why would you even think thoughts like those?
<gchristensen>
y'all, rusts' readdir iterator absolutely does continue to return new files, but not deterministically
<abathur>
I'm not saying I'm compiling a list of modern curses, but if I was, "may your iterators return nondeterministically" might get a spot
<Church->
gchristensen: Yep, it's not to my knowledge.
<Church->
Hmm, borked my teleport module somehow... now what did I do...
iqubic has joined #nixos-chat
<gchristensen>
blah. it is annoying that the inotify crate requires this &mut buffer
buckley310 has joined #nixos-chat
<Church->
Writing a watcher daemon of a sorts?
<gchristensen>
yea
<abathur>
anyone know if there's a good way to flaunt the point of the standard and target opengraph tags a certain platform to discourage sharing there? :)
<abathur>
s/a/at
<abathur>
oh, at a, even; must be that time of night
<colemickens>
"hook reply is 'decline'" oh what I would give for a reason
<colemickens>
`sudo ssh user@bldr` works, `user` is in trustedUsers, that should be it?
<colemickens>
(it's even marked big-parallel)
endformationage has quit [Quit: WeeChat 2.7.1]
<Ashy>
ooh this looks nice: 15.17.3.6. How to consume python modules using pip in a virtual environment like I am used to on other Operating Systems? from https://nixos.org/nixpkgs/manual/#python
<Ashy>
didnt realise venvDir was an option to automatically do the venv step
<Ashy>
side note, why isnt there a direct anchor link to every subheading in the nix manuals?
<colemickens>
+1
cjpbirkbeck has quit [Quit: Goodbye, take care]
* colemickens
is beyond frustrated
* colemickens
seriously needs some hope of an alternative nix implementation
<Church->
Heh oh?
<Church->
I mean there is I think? Haskell and rust right?
drakonis has quit [Ping timeout: 260 seconds]
<sphalerite>
Church-: not of the daemon
<Church->
Ah, point
<colemickens>
I'm a bit frustrated. I'm really disappointed that I am once again losing hour+ to wondering why nix won't pick my remote builder
<sphalerite>
also not sure how complete the evaluation side of hnix and rnix are
drakonis has joined #nixos-chat
<sphalerite>
and I'm pretty sure rnix can't talk to a store in any way, while hnix-store exists but isn't complete I think
* colemickens
now 90 minutes in on trying to figure out why nix won't use my builder. again. for the 6th+ time in using nix
iqubic` has joined #nixos-chat
iqubic` has quit [Remote host closed the connection]
iqubic has quit [Ping timeout: 260 seconds]
<sphalerite>
colemickens: for all derivations, or just a specific one?
<colemickens>
sphalerite: it's a system derivation that I happen to be building.
<colemickens>
I can test with a few smaller ones to spot check in a second, that's a good idea I should've done
<sphalerite>
colemickens: oh, those have preferLocalBuild I think.
<colemickens>
I have another suspicion in flight.
<sphalerite>
colemickens: if it's indeed a nixos system toplevel derivation.
<colemickens>
(this is a build I do regularly, just a new builder machine)
kalbasit has quit [Ping timeout: 264 seconds]
<colemickens>
I do remotely, regularly, even :P
<colemickens>
I think that I've actually somehow managed to remove root from trustedUsers.
<sphalerite>
colemickens: also, what really irritated me in the past and cost me quite a bit of time is that you need to enable nix.distributedBuilds as well as setting nix.buildMachines on nixos.
<colemickens>
And I'm wondering if its related
<sphalerite>
though I guess if you've been using other builders that's not hte issue.
<colemickens>
sphalerite: luckily... luckily I think I avoid that with my strong reluctance to touch nix.conf at all, even for my cachix mirrors
<colemickens>
but yeah, I did actually see that one mentioned in the middle of the wiki page about an hour ago :) woo
<sphalerite>
how are you doing it then? Passing --builders?
<colemickens>
yeah
drakonis has quit [Quit: WeeChat 2.8]
<sphalerite>
And how are you specifying the builder?
<Arahael>
eyJhb: Against apple's T&C's of using beta software.
<Arahael>
(Usually)
<eyJhb>
Ahh, the company of BS
<eyJhb>
Lately I have been thinking about making a FDroid repo using NixOS to build the APKs and generate the files in general. Anybody know if someone is currently doing this?
<{^_^}>
tadfisher/gradle2nix#13 (by pstn, 13 weeks ago, open): Feature Request: Make gradle2nix work for android projects
<philipp[m]>
But I think nealry every android user here would be so down for that.
<bqv>
i was actually trying to use that, on sunday
<bqv>
didn't realise it didn't work
<bqv>
explains a lot
<eyJhb>
bqv: Did you spend waaay too much time on it?
<eyJhb>
philipp[m]: It would be coool! Also just to have your own patched APKs.
<bqv>
considering the progress i made, yes
<eyJhb>
Not that I ever patch my banking app
<philipp[m]>
Yeah, but patching/upgrading newpipe in time would be nice!
<philipp[m]>
Or building signal yourself and putting it in a repo.
<eyJhb>
The worst is the signing part I guess
<eyJhb>
I have a small dream, of setting up a server, putting a service on that pulls from a repo, and then throw the SSH key away
<eyJhb>
So that it can just run on its own :p It is stupid, but could be fun
cjpbirkbeck has joined #nixos-chat
__monty__ has joined #nixos-chat
<philipp[m]>
Yeah, sounds fun.
<eyJhb>
Then it would just be the git repo, that controlled what would be in the fdroid repo
<philipp[m]>
Yeah, signing could be tough to do in nix.
<eyJhb>
In general it is for FDroid
<philipp[m]>
Huh? I thought the problems would stem from not being able to have secrets in the nix store?
waleee-cl has joined #nixos-chat
<eyJhb>
read the secret from a file?
<eyJhb>
Also, what is the site called, that has a count of packages in different repos?
<__monty__>
repology
<eyJhb>
We are so close, to having more packages than AUR :(
<philipp[m]>
Shouldn't that be blocked by the sandbox?
parsley936 has joined #nixos-chat
<eyJhb>
I should really really be doing maths...
<eyJhb>
While building? Yes I guess so
<adisbladis>
eyJhb: I think we already do
<adisbladis>
Repology stats is not counting everything iirc
<adisbladis>
Emacs packages are excluded for example
<eyJhb>
Yay Nix!
<eyJhb>
adisbladis: are you still supporting vgo2nix btw?
<adisbladis>
eyJhb: Well, let's pm ;)
<eyJhb>
Also, do we have a real count?
bqv has quit [Quit: WeeChat 2.9]
<philipp[m]>
How would we even count? Do we just count all the packages and have every python lib 9 times?
<adisbladis>
It's really hard to count
<adisbladis>
Also because we consider a lot more things as "packages"
<adisbladis>
runCommand <- also returns a derivation
<eyJhb>
EVERYTHING :p
<eyJhb>
philipp[m]: yes x9
<eyJhb>
THat is just how it goes
<adisbladis>
Need to inflate those numbers
<philipp[m]>
Ahahahaha! For some... unfortunate verbosity reasons we just discovered that journald has no problem in writing 5M entries in 10 Minutes. Log retention time is kinda suboptimal though.
<philipp[m]>
We are talking about a small vm on spinning rust here.
andi- has quit [Ping timeout: 246 seconds]
andi- has joined #nixos-chat
karolus has joined #nixos-chat
<eyJhb>
philipp[m]: better than a car spinning rust
drakonis1 has quit [Ping timeout: 256 seconds]
drakonis1 has joined #nixos-chat
<eyJhb>
philipp[m]: but is Gradle really required for everything?
karolus_ has joined #nixos-chat
<eyJhb>
Damn it
<eyJhb>
All those withou ZNC, it is a love/hate
<eyJhb>
I can see you are offline, but I can't really leave messages
<philipp[m]>
"eyJhb" (https://matrix.to/#/@freenode_eyJhb:matrix.org): not everything but enough to matter in practical applications, I think.
<eyJhb>
Holy hell, why did it prefix your message with the worlds longest string?
<{^_^}>
vector-im/riotX-android#1506 (by f0x52, 5 weeks ago, open): Badly formatted mentions in body
<eyJhb>
Valodim: didn't they release a successor?
<eyJhb>
Or was that RiotX
<Valodim>
that is riotx
<infinisil>
Random question: I'm ordering something from canada (a sticker), which is really cheap (even has a discount), but I want to support the artist by leaving a tip, which I can choose myself
<infinisil>
I enter $10, but then it complains that the amount should be at most the price of the item itself. But the payment option isn't disabled with that amount
<infinisil>
So, why would somebody limit the tip to the price of the item itself? Is this something about the law that they could get in trouble with?
<infinisil>
(as in, if they get more than 100% tip)
<eyJhb>
Well, I would assume tips are usually some % of the price
<eyJhb>
But not sure why it would be limited
<philipp[m]>
riotx was renamed element for the first live release. Also sorry for messing with you irc peeps. I'll work around it until it is fixed.
<eyJhb>
Question, if you deconstruct something in reverse, then.. Are you then constructing it?
<eyJhb>
Also, COME ON gradle2nix, this takes forever
<eyJhb>
<=============> 100% CONFIGURING [24m 23s], where it has spent the 20 minutes being 100%
<infinisil>
I just sent a message asking for clarification, will report back when I get a reply :)
drakonis1 has quit [Read error: Connection reset by peer]
<__monty__>
Worrisome Wednesday, when CVEs roam the prairy.
<infinisil>
cal <3
<eyJhb>
However hate that it defaults to day 0 = sunday
<bqv>
just work modulo 8 and start numbers from 1
<bqv>
problem solved
<gchristensen>
eyJhb: right? monday should be 0
<infinisil>
alias cal="cal -m"
drakonis2 has quit [Read error: Connection reset by peer]
drakonis2 has joined #nixos-chat
<eyJhb>
Yeah I considered that infinisil , but it is just.. Why
<eyJhb>
The same when stuff defaults to imperial :|
drakonis_ has quit [Ping timeout: 272 seconds]
<infinisil>
Yeah
<infinisil>
I'm often on reddit, where most people assume everybody is american, so I frequently need to convert units..
<eyJhb>
There is a reason, I always convert to EUR when I can etc. :p Forcing Americans to take other currency into consideration, same with always using metric :D
drakonis2 has quit [Client Quit]
<eyJhb>
But I think there is the convert bot on reddit infinisil ?
<gchristensen>
infinisil: aliases are for novices. No, we have Nix: self: super: { utillinux = super.utillinux.overrid.... }
<bqv>
assuming currency in any situation seems ":/"
<infinisil>
I don't mind dollars tbh, currency is something almost inherently diverse, and the US dollar is like a standard almost
<infinisil>
gchristensen: Hehe yes
<bqv>
i quite intentionally don't ever convert anything to imperial units
julm has quit [Quit: leaving]
julm has joined #nixos-chat
<infinisil>
bqv: I intentionally put both so americans can get used to metric :)
<eyJhb>
infinisil: But...
<eyJhb>
Will they ever learn?
<bqv>
nice principle
<infinisil>
In due time I hope
<bqv>
but i really just don't want to feed into the americentricism of the internet
<infinisil>
I mean we can't bunch everybody together, there's many americans who do use metric already
<__monty__>
Hasn't the US officially adopted the metric system? It's just the people being stubborn, right?
<eyJhb>
__monty__: it is the US being stubborn
<eyJhb>
They offically use it
<eyJhb>
But don't
<eyJhb>
Also the imperic meassurements are set based on the metric system
<eyJhb>
(defined by)
<infinisil>
I have to admit, it's not easy to get used to another measurement unit
<eyJhb>
So basically, not until the schools switch I would say
<eyJhb>
But bqv americantricism is much more than this :p
<infinisil>
And maybe with america becoming more.. scientific
<eyJhb>
But I don't think that is NixOS allowed
<gchristensen>
my house used celsius for a while but it wasn't so good when I moved to the frigid north
<__monty__>
I can never remember the F to C ratio.
<gchristensen>
so
<infinisil>
"However, on September 23, 1999, communication with the spacecraft was lost as the spacecraft went into orbital insertion, due to ground-based computer software which produced output in non-SI units of pound-force seconds (lbf·s) instead of the SI units of newton-seconds (N·s) specified in the contract between NASA and Lockheed."
<gchristensen>
infinisil
<gchristensen>
apparently
<gchristensen>
that was caused by a feud between two specific people on a specific team who refused to talk to each other, and intentionally antagonized eacho ther
<bqv>
__monty__: add 32, divide by 5, multiply by 9, or something?
<bqv>
alternatively, use a popular search engine
<eyJhb>
I will never get scientist, not using SI units/the metric system
<__monty__>
bqv: That sounds like C to F.
<bqv>
oh, it does
<bqv>
eyJhb: but, kelvin!
<infinisil>
gchristensen: Interesting
<eyJhb>
kelvin is a SI unit bqv
<eyJhb>
:p
<infinisil>
gchristensen: I need to ask: Got a source for that? (can't find it in wikipedia)
<bqv>
oh, it is
<infinisil>
(unless I missed it in my quick look-through)
<eyJhb>
I think one of the funniest things, is when the radio apologies for playing the US version of songs :p (not that this should turn into a US hating thing)
<gchristensen>
infinisil: unfortunately the source isn't wikipedia grade: growing up with a lot of nasa and lockheed martin employees in my life :)
<eyJhb>
Of course it is :D It is the standard way to measure temp in science I would say. At least it was pretty much the only thing we used we calculating stuff
<eyJhb>
gchristensen: NixOS in space.
<gchristensen>
eyJhb: yeah I'm still trying to get them to say so publicly :(
<infinisil>
gchristensen: Oh cool
<eyJhb>
That they use NixOS?
<bqv>
my headphones have such a nice mac address... 11:11:22:33:33:98
<eyJhb>
That..
<eyJhb>
BT?
<bqv>
yes
<eyJhb>
Makes more sense, but still
<eyJhb>
The hell
<infinisil>
eyJhb: I think kelvin has to be used in many calculations because it's the unit where 0 actually means 0
<eyJhb>
Yuuup
<eyJhb>
273.15 + \degree C
<eyJhb>
As far as I remember
<infinisil>
Yee
<eyJhb>
Not valid in the form of units
* infinisil
once made a unit library and temperatures were annoying
<bqv>
wonderful, headphones are go
<eyJhb>
Why infinisil ?
<eyJhb>
bqv: go?
<bqv>
working
<infinisil>
Just wanted to, not sure if there was any other motivation :P
<eyJhb>
Do.. Do you hate yourself sometimes?
<eyJhb>
\s :D
<infinisil>
It was in Swift, which was very new at the time, and I don't think there was a good unit library already
<infinisil>
Other than temperatures, unit libraries are pretty nice to write!
<bqv>
lol
<eyJhb>
Why did temp suck so much?
<infinisil>
It's the only thing where you can't just multiply between different units to get others
<infinisil>
kelvin -> celsius needs +
<infinisil>
celsius -> fahrenheit needs + and *
<eyJhb>
No no
<eyJhb>
You just leave out F, and throw a exception infinisil
<infinisil>
Perfect lol
<gchristensen>
raise "temperature is a lie anyway"
<eyJhb>
ERADICATE
<infinisil>
There might be more units like that though
<joepie91>
SensibilityException: Why are you using Fahrenheit
<eyJhb>
It is a man made construct?!
<gchristensen>
I mean you can't do meaningful things with it like ... division
<eyJhb>
joepie91: sounds like a ex
cole-h has quit [Quit: Goodbye]
<bqv>
joepie91: lmao
<eyJhb>
SOME might be mad about it however
<eyJhb>
'MURICA
<bqv>
I feel like i really shouldn't be involved in this conversation... as a brit, i'm genetically engineered to hate americans, and holding that back is challenging
<ajs124>
I mean, as a brit, you also have some cool and good units. Like stones and pints.
<eyJhb>
ajs124: fucking miles
<ajs124>
nautical or imperial?
lovesegfault has quit [Quit: WeeChat 2.8]
<eyJhb>
imperical
<bqv>
i really wish we'd drop miles... i'm pretty sure stones is reserved for angry old men in pubs, though
<bqv>
and i hadn't thought about pints, that's unfortunate too
<eyJhb>
Don't you use lbs as well?
<bqv>
again, angry old men
<bqv>
i meet very few people who don't prefer kgs
<ajs124>
psi?
<eyJhb>
Looks smaller on the weight
<bqv>
i'm no meteorologist
<bqv>
can't say
<eyJhb>
We have used PSI waaay too much at times
<ajs124>
When it comes to "exotic" units, most of the planet uses some dumb stuff. Like AWG.
<eyJhb>
Awg?
<eyJhb>
Ah
<bqv>
oh, audio's not working
<bqv>
that's not great
<bqv>
i think i know why, though
<eyJhb>
The most weird units I have used, is when my class visited Rome and had to measure a building with what we had on us. We measured the hell out of it, using 2.5 shoes, 3.8 hugs, etc.
<bqv>
nevermind it was just muted..
<bqv>
hugs?
<bqv>
whaat on earth
<eyJhb>
We had to measure the pillars
<eyJhb>
So hugs was the best way to do so
<bqv>
lol
<eyJhb>
We basically had a napkin home saying stuff like the above, '3.8 Lasses hugs'
<eyJhb>
'3.8 <name> hugs'
<ajs124>
Also hardness. There's so many scales. Rockwell, Shore, Mohs, …
<samueldr>
other parts of the commonwealth still live with a weird hybrid system
<samueldr>
here in Canada you're more likely to know your weight in pounds, and height in feetsies
<samueldr>
buy meat/vegetables in pounds
<joepie91>
feetsies
<samueldr>
but everything else metric
<bqv>
yikes
<samueldr>
including temp
<joepie91>
feetsies.
<samueldr>
lol
<bqv>
i am 1.76m and 81kg
<samueldr>
I had to 2.2 the second one, to get a rough idea
<samueldr>
to ×2.2*
<joepie91>
'pond' and 'ons' are fairly common units here for buying weighted groceries at eg. a market
<__monty__>
Dutch people are weird.
<__monty__>
samueldr: General commonwealth or just US-adjacent commonwealth?
<samueldr>
oh, just saying: it's in advertised prices that they use pounds, but the transaction is in kilos
<joepie91>
can confirm
<joepie91>
but only as whole integers, and anything in a supermarket is gonna be labelled in (kilo)grams
<samueldr>
__monty__: I think at least the UK is part of the commonwealth, and not really US-adjacent
<joepie91>
advertised prices here are always per kilogram
<joepie91>
or per prepacked amount
<samueldr>
yeah, actual items are labeled in kilos, the advertised price is in pounds
<samueldr>
so you buy 1.394kg of the meat advertised at 2.99/lb (some other price in smaller letters /kg)
<bqv>
zuckerberg looks even weirder than i remember
<__monty__>
Weird.
<ajs124>
around here you can also order beer in Maß, which by now is just a fancy word for 1 liter
<ajs124>
and order a Halbe if you only want 500ml of beer
<joepie91>
bqv: how many % Data is he by now
<__monty__>
Gib mir más Bier!
<bqv>
suffice to say, mostly
<joepie91>
lol
<samueldr>
to finish about price in pounds, it's all about showing cheaper prices
<__monty__>
What about gallons though?
<samueldr>
now, about people's measurements? customs
<samueldr>
that's crazy talk
<samueldr>
milliliters and liters all the way
<joepie91>
samueldr: oh I was actually wrong when I said that everything is priced in kilograms here
<samueldr>
!
<__monty__>
Gallon of gas?
<samueldr>
liters
<drakonis>
aight people, who's up for playing tabletops?
<joepie91>
samueldr: for more expensive stuff, like the more expensive cheese, it is often priced per 100gr to make it look cheaper :)
abathur has joined #nixos-chat
<samueldr>
ah
<drakonis>
i have acquired myself a copy of foundry
<samueldr>
that's still "by the [kilo]gram" at least
<joepie91>
aye
<joepie91>
samueldr: one thing I like is that in supermarkets in NL, there's basically always a per-<unit> price on price tags
<joepie91>
I suspect it's legally mandated
<joepie91>
makes it easy to compare prices between stuff in different packaging sizes
<joepie91>
so per kg, per liter, etc.
<samueldr>
joepie91: we have this too
<samueldr>
provincially mandated
<samueldr>
anything food-related MUST have a per-unit price
<samueldr>
though "unit" is loosely enough defined
<samueldr>
I don't remember which product, but something that could be actually sold both by weight and volume had labels with different units
<samueldr>
here I assume absolutely no malice :)
<joepie91>
yeah I've occasionally seen that here also
<joepie91>
:P
<bqv>
oh shit
<__monty__>
How much are spoons? 27 cents per gram. Ah, thanks.
<bqv>
zuck's been cornered by the congressman
<joepie91>
samueldr occasionally you get the most asshole-y move of putting a per-kg price on the bulk packaging and a per-piece price on the individually-packaged ones
<joepie91>
bqv: deets
<joepie91>
__monty__: are your spoons made of silver
<joepie91>
because those are some expensive damn spoons
<joepie91>
:P
<__monty__>
No, lead spoons only. That way you exercise while eating.
<__monty__>
And the heavy metal poisoning also helps cutting those last few lbs.
<bqv>
gradle is definitely a mistake, i was doing my darndest to avoid it. you might be able to avoid it quite easily depending on what you're doing
<eyJhb>
I want to package Android apps in general
<eyJhb>
And.. Most use Gradle
<bqv>
house rep: "does apple have full control on whether an app gets onto the app store" tim cook: "well, if it's a native app, yes, if it's a web app..." house rep: "..yeah ok thanks"
<samueldr>
gradle is make to gradle-ly drive you insane
<samueldr>
made to*
<eyJhb>
Get. Out.
<eyJhb>
:D
<eyJhb>
You just have them loaded don't ya samueldr ?
<samueldr>
I bat them as they come
<eyJhb>
I need a Gradle expert
<bqv>
i'm loving this committee, they're really holding these companies over the fire
<eyJhb>
Time to force feed NixOS to some Gradle users
<eyJhb>
Never had to give my cat pills, but she loved the last medicin I get for her
<eyJhb>
got*
<eyJhb>
Which was just.. Get her to drink this
<eyJhb>
Not sure they will take my NixOS question
<bqv>
oh snap, google work with the chinese military, but not the american military
<bqv>
nevermind, fake news
<eyJhb>
TRUMP ^
<drakonis>
google works with everyone
<drakonis>
whoever makes the most money
lovesegfault has joined #nixos-chat
Church- has quit [Ping timeout: 264 seconds]
aleph- has joined #nixos-chat
<bqv>
are sergei and larry not in charge at all anymore?
rajivr has quit [Quit: Connection closed for inactivity]
<danderson>
nope
<danderson>
sundar's CEO of both Google and Alphabet. L&S are still on the Alphabet board, and I think are still employees thereof
<danderson>
but no longer in the chain of command, as of early this year
<__monty__>
Guess their personal values were in the way.
<bqv>
interesting
<bqv>
yeah it sounds like it
<danderson>
it's... unclear.
<danderson>
there's two major plausible explanations, IMO
<bqv>
a congresswoman described how L&S had as part of their personal philosophies that users would never be linked with cross site cookies, but under sundar and with the acquisition of doubleclick they now are
<danderson>
one is they suddenly realized they were megarich and weren't getting paid enough to deal with google/alphabet's reckoning with harboring predators, and the ruckus around google getting into the weapons business
<danderson>
so just decided to peace out
<danderson>
the other is that their leaving coincides quite well with a bunch of other revelations of higher-ups at google being sociopathic predators on their employees, who were repeatedly protected by executive leadership.
<danderson>
so maybe they were "helped" out because of that, or bailed ahead of other stuff going public
<danderson>
fwiw, doubleclick specifically is still a really weird silo within google. There are a lot of restrictions on cross-pollination between that information and other google data.
<danderson>
whether that's enough, or has changed since the original agreements, or what... I don't know.
<__monty__>
In the hearing sundar said it was no longer siloed though.
<danderson>
well, there ya go, my info's outdated.
<__monty__>
Or, he didn't deny when the congresswoman asked him.
<danderson>
that said, I don't think that was down to L&S having philosophical objections or anything like that. If anything, I'd expect eric schmidt to have that kind of thoughts.
<danderson>
I expect the google myth is that L&S were wonderful principled leaders, and it's going out the window now with sundar. Having been on the inside, that's really not how it went down.
evanjs has joined #nixos-chat
<danderson>
sundar's definitely a more "utilitarian" leader than L&S, in the sense that it's all about maximizing board happiness and revenue.
<danderson>
but that doesn't mean L&S were *that* opposed to that either.
<danderson>
it's more a continuation of how they ran things, than completely new management.
<__monty__>
All you need to do to confirm that is look at stock prices.
waleee-cl has joined #nixos-chat
<bqv>
oh man
<bqv>
zuck's squirming again
<samueldr>
where o where has the falsehood that `bs=some block size` make `dd` invocations fail, or slower even started?
<samueldr>
always touted around and parroted without batting an eye
<samueldr>
there's also the "setting a bs will make dd go slower since otherwise it will detect it" which is patently false accordin to the manual AND the source :<
<samueldr>
and all the arguments about that make no sense considering that they all assume that dd *somehow* goes around the block device drivers of the OS
<samueldr>
paraphrased "flash has lower sustained write speed so the flash may fail"
<samueldr>
like, I'm fine with learning I'm wrong
<bqv>
ah, so it wasn't fake news, google does work with the chinese military
<samueldr>
I even held that non-fact-based cargo-culted opinion at some point in the past
<samueldr>
but I looked at things and read about it all and I don't see how it makes sense
<samueldr>
basically, what I'm asking is "I don't think I'm wrong, but I want to know if I am"
<ajs124>
if you've read the manual and source code and they both agree with you, odds are you're right
<eyJhb>
Src > Manual
<samueldr>
somehow that's not enough for the other individual :/
<samueldr>
eyJhb: agreed :)
<ajs124>
how are they backing up their opinion? measurements? anything?
<samueldr>
"experience"
<samueldr>
I'm either a noob or lucky
<ajs124>
ah, so "nothing". ok.
<samueldr>
that's when I said "I'm off"
<samueldr>
serves nothing to argue on dogma
<__monty__>
samueldr: Does it end up using smaller buffers if you specify a block size?
<__monty__>
Well, smaller or larger I guess, depending on what block size you'd pass.
<samueldr>
when you don't it's as if you said bs=512
<samueldr>
because that's what the source does
<samueldr>
so "not setting one" is "using the default 512"
<samueldr>
and speed-wise, my *experience*, since it's important, tells me it matters a bunch when writing
<samueldr>
and I can't see how it can somehow fail if it's too big
<samueldr>
in my experience too big has only made the speed plateau
<ajs124>
what might be happening is that you're reading source and documentation of one dd version/implementation but they made their experiences with another. like, gnu vs. bsd vs. busybox vs. whatever.
<samueldr>
possibly
<samueldr>
that's an assumption I was keeping at the back of my head
<samueldr>
but nothing from those different individuals end up showing me that they actually have knowledge
<samueldr>
and that it's not a cargo-culted idea
<ajs124>
like that gnu tar thing, where it tries to access files over rsh. nobody would agree that that's something tar does and you can probably find 8 implementations that don't but gnu tar, in all its gnu wisdom does.
<samueldr>
I'm also open to reading docs about that, because I can't even find mentions online about that misconception
<samueldr>
no "bs= considered harmful" drivel
<__monty__>
I've taken to staying away from dd unless I have a good reason. (Which usually means nothing else worked.)
<samueldr>
heh
<ajs124>
why?
<eyJhb>
__monty__: dd if=/dev/zero of=/dev/sd*
<eyJhb>
?
<eyJhb>
:D
<infinisil>
I keep wondering how new nicks in #nixos know about ,channels
<__monty__>
Don't want to deal with blocksizes and stuff. The defaults in cp and rsync for example are good enough for me.
<gchristensen>
infinisil: /topic
<infinisil>
Oh!
<ajs124>
right, but you can't cp and rsync contents of one blockdevice to another, can you now? ^^
<infinisil>
It all makes sense now
<samueldr>
gchristensen: impossible, nobody reads the topic ;)
<gchristensen>
a new nick who comes in and does ,channels my respect starts out 15 point higher
<philipp[m]>
Do they change their from? Because I'd just blacklist it with sieve or sth.
<__monty__>
Similar to the dd thing. Both bsdtar and gnutar have supported autodetection of compression by the extension since practically forever. Yet because the posix man page doesn't say so some usages are less compatible across platforms for the sake of "compatibility."
<samueldr>
dunno, maybe I am naïve, but I would expect things not be broken left and right forever and ever
<ashkitten>
nice, just tested watching a video i knew suffers a lot from the cpu being pinned, side by side two instances of firefox one with vaapi and the other without
<ashkitten>
(while the cpu was being pinned with mprime)
<ashkitten>
the one without vaapi enabled was frozen 99% of the time,
<ashkitten>
the one with it enabled worked very smoothly
<__monty__>
Is it enabled by default?
<ashkitten>
no
<ashkitten>
it only works on wayland with webrender and wayland-dmabuf enabled
<ashkitten>
and for vp9 you need to disable ffvpx
<aaronjanse>
Huh. I thought that the best way to write to an image is simply `cat source > dest.iso` because the block size etc are auto chosen
<aaronjanse>
Along with the potential for viewing progress with `cat source | pv > dest.iso`
<aaronjanse>
Oh crap
<aaronjanse>
Would `< source > dest.iso` work? I think so
<samueldr>
aaronjanse: yeah, though here it was about `dd` specifically
<aaronjanse>
Oh got it
<samueldr>
I know that "other methods' perfs will vary"
<aaronjanse>
Ah
<samueldr>
though other methods don't get oflags=direct,sync :)
<__monty__>
Does cat detect block size? Or does it just use a several MB default?
AluisioASG has quit [Ping timeout: 256 seconds]
<samueldr>
it ends up filling the buffer, and the kernel just writes
<samueldr>
AFAIUI
<samueldr>
so it should go as fast as it can
<__monty__>
Don't cp and rsync do essentially the same thing?
<ldlework>
cp doesn't do deltas does it?
<__monty__>
Neither does cat?
<bqv>
i once wanted to train myself to always use rsync over cp
<bqv>
since rsync is effectively a strict superset, and still as ubiquitous
<bqv>
joepie91: oh, i wanted to ask you, what would you reccomend over JWTs for authentication?
<bqv>
i recall you had issues with JWT use
<__monty__>
I do that. You pay in performance for local transfers but I've found whenever I care about that it's because the medium's inherently slow and I'd far rather have --partial in case the transfer's interrupted.
<joepie91>
bqv: yes, JWT is terrible. but to make a better recommendation I need more info about the usecase :P
<joepie91>
__monty__: I've actually often had rsync be faster than cp for me
<bqv>
joepie91: what might be options?
<joepie91>
for cross-device copies
<__monty__>
Incremental?
<joepie91>
bqv: well, the main thing I'm looking for is whether you have a session usecase (website/webapp that works in a browser), a private API usecase, a public API usecase, or some weird setup with statelessly scalable servers for downloads or whatever
<__monty__>
Because that doesn't make sense to me for from scratch transfers. rsync needs to do a lot more checks, no?
<joepie91>
__monty__: nope, full
<joepie91>
__monty__: mainly for folders with many small files
<__monty__>
Any special options you're passing?
<joepie91>
something that rsync is doing seems to work very well for small files
<joepie91>
on HDDs anyway
<bqv>
joepie91: essentially just session usecase, alone
<joepie91>
the usual -avz --progress
<__monty__>
Hmm, interesting.
<joepie91>
bqv: okay, then you want plain ol' session cookies. what language and/or framework are you using server-side?
<bqv>
session cookies aren't stateless, would it not be worth using a jwt purely for encoding sessions?
<bqv>
i assumed you had a better option
<bqv>
i'd be using some asp variant
<joepie91>
bqv: sessions in general are not stateless
<joepie91>
as in, sessions as a concept are inherently stateful, "making them stateless" is not an existent option
<joepie91>
because to implement them securely, you need revocation
<joepie91>
(!= expiry)
<bqv>
that's a fair point, tbh
<joepie91>
the only reason people can do "stateless" sessions with JWT is because they just pretend that revocation doesn't exist and isn't needed, basically :P
<joepie91>
there's a bunch of people who try to tack revocation onto it afterwards, and they of course end up with a stateful system, but then claim it is still stateless because it's JWT
<bqv>
yeah, that i've seen
<joepie91>
bqv: aside from that, "stateless sessions" also just... aren't very useful
<joepie91>
like, there's two decades of tools and approaches for scaling stateful session cookies, which work perfectly fine, sites like Reddit and Facebook still use them
<joepie91>
it's mostly a solution looking for a problem
<bqv>
mm, i was mostly thinking it saves on computation
<bqv>
but fair enough
<joepie91>
sure, but nowhere near as much as people like to claim
<joepie91>
consider that session data retrieval is basically by definition an index lookup, ie. the fastest DB query you can possibly have short of a literal `SELECT 1;`
<joepie91>
if you were to profile your application, almost any application, it is incredibly unlikely that session handling overhead will show up anywhere near the top of the list
<bqv>
yeah
<joepie91>
it's non-zero, sure, just not non-zero enough to matter :D
<bqv>
understandable, i'll go with that
* colemickens
wonders how many auth frontends keep a cache of already crypto-verified JWTs too
<joepie91>
oh yeah, that's a funny one actually
<joepie91>
if you don't have hardware crypto acceleration - which especially many VMs don't - signed tokens might actually cost you more CPU than that stateful DB lookup :D
<bqv>
lol
<joepie91>
(compared to unsigned session cookies)
<samueldr>
colemickens: tavis' opinion is founded under the premise that _for security_ you don't _need_ reproducible builds
<samueldr>
colemickens: but to me, reproducible builds for security is only one leg of it, and probably the least important
<samueldr>
(to me)
<samueldr>
correctness!
<samueldr>
reproducibility is important to show that your toolchain produces the expected results compared to another one
<colemickens>
Oh yeah, for sure. I think it's sort of a weird article, but I agree with you and I think some of hte other commenters
<samueldr>
I don't even think Tavis thinks it's not good, but it's overhyped surely
<colemickens>
I just meant the folks opining about Dockerfiles and reproducability :)
<samueldr>
oh
<samueldr>
no
<gchristensen>
hahaha.
<bqv>
my god, even the Table API on cosmosdb doesn't enforce a scheme
<andi->
anyone aware for >FHD 13-14" notebooks with AMD graphics and no intel or NVIDIA shit in them?
<samueldr>
oh, ASUS has one (no, not the G14 with intel) I think
* samueldr
has no idea how to coerce the asus website in showing it to me
<andi->
let me extend the exlcude list: no weird screen on the touchpad :D
<samueldr>
haha
<samueldr>
I would love one of those to play around with the feature
<joepie91>
as the built-in will likely be guaranteed, to, uh
<joepie91>
....meet the quality level of the rest of the tools
<bqv>
tbh i'm mostly asking this in terms of personal research, i wanted to know what alternatives are about
<joepie91>
right
<joepie91>
bqv: so more hypothetically, the main other two options are API tokens and non-shitty stateless tokens
<bqv>
the former being just a form of stateful role-based auth, and the latter based on jwts?
<joepie91>
for externally consumable APIs, for example, you'd usually want to use API tokens, literally just randomly generated tokens (usually an identifier / secret pair, for timing attack reasons)
<joepie91>
saves you the overhead and additional attack surface of the crypto which you don't need anyway for that case
<joepie91>
sometimes you really do need stateless tokens as strictly single-use transferable claims, notably in cases like SSO services (single use, just to get a session cookie elsewhere) and separately-scalable services (transcoding clusters, file download servers, etc.) -- in those cases, PASETO is a much safer option
<joepie91>
definitely not based on JWT
<joepie91>
because its cryptographic design is complete junk
<samueldr>
joepie91: but it has JSON, and JS in it, so it's webscale
<joepie91>
there's a reason why basically every JWT implementation in existence has suffered, at some point, from a vuln that allowed a complete auth bypass
<bqv>
someone make Rust Web Tokens!
<joepie91>
it's a direct result of the shit spec
<bqv>
i recall
<joepie91>
samueldr: huh? definitely does not have JS in it
<bqv>
JSON has JS in it
<samueldr>
(that's a joke, sorry it didn't land)
<bqv>
transitively true
<samueldr>
(because we know that webscale is a joke)
<joepie91>
bqv: anyway re: API tokens, exactly what access is granted is up to your app, I'm just talking about the practice of randomly-generated stateful tokens :P
<joepie91>
samueldr: oh right :P
<andi->
all my mongodbs disagree ;-)
<bqv>
joepie91: yeah, that's kinda what i had in mind r.e. alternatives
<bqv>
fair enough
<samueldr>
andi-: meow?
<bqv>
andi-: woof!
<andi->
hrhrhr
<samueldr>
(I don't remember which not-a-db database thingy the "meow" attack attacks)
<joepie91>
bqv: but yeah, the recurring theme here is "do not add crypto unless you actually need it, and if you do, pick the crypto built by someone actually competent" :D
<joepie91>
samueldr: all of them
<samueldr>
nice
<joepie91>
(fairly literally, I believe)
<bqv>
yup
<__monty__>
joepie91: Aren't SSL implementations plagued by that same caveat?
<andi->
the internet is not for private/confidential/trustworthy data. Things get so much easier :-)
<joepie91>
__monty__: similar yeah
<joepie91>
andi-: I think that ship has sailed the world a few times by now
<joepie91>
:P
<andi->
oh noes :'(
<philipp[m]>
I mean there were those people that tried to go the way of "post privacy"...
<philipp[m]>
It didn't go particularly well, I'd say.
<bqv>
post privacy?
<bqv>
what?
<philipp[m]>
There was a little movement in German digital politics sub cultures that argued to just get rid of privacy.
<philipp[m]>
The movement pretty much died when the snowden leaks started.
<__monty__>
Isn't privacy essential for democratic government? (Not that today's governements are all that democratic.)
<philipp[m]>
There are about a million problems with post privacy. Probably more.
<samueldr>
but privacy ain't one
<joepie91>
philipp[m]: just going out on a limb here, did the proponents of said "post privacy" ideology happen to be primarily male, white, native German?
<philipp[m]>
You forgot moderately wealthy, well educated and from intact families.
<joepie91>
wasn't familiar enough with conservative German culture to make those specific assumptions :)
<joepie91>
but they also don't surprise me
<joepie91>
I've seen the post-privacy ideology before, in various places, it always happens to be espoused by those with a powerful position in society...
<bqv>
huh, representative jamie raskin
<bqv>
MichaelRaskin: you have a namesake
<philipp[m]>
I wouldn't say they were conservative per se, but damn was there a lot of unreflected privilege to unpack...
<joepie91>
philipp[m]: oh yeah, not saying they would need to be consciously conservative. rather that the socioeconomic position of these people always happens to align with whatever conservative ideology in a country centers arond
<joepie91>
around*
<__monty__>
The "Privacy's too hard, why even bother?" sentiment's pretty common outside those circles though.
<__monty__>
Usually specifically online/digital privacy.
<joepie91>
__monty__: there's a difference between not caring about your own privacy, vs. thinking that others shouldn't
<joepie91>
the latter is definitely very specific to demographics that overlap with conservative ideology
<joepie91>
those with privilege in a given society
<joepie91>
which is unsurprising, because if people actually understood and recognized the situation of unprivileged demographics, they would not be arguing for it :)
<__monty__>
I assumed these post-privacy proponents were leading by example.
<joepie91>
doesn't really matter, my point is rather that to those with a privileged position in society, privacy may not seem like that big a deal because they are not under threat
<philipp[m]>
A few of them actually were. One published his gps coordinates 24/7.
<philipp[m]>
but yes, what joepie91 says.
<joepie91>
it's just another form of "I don't personally care about this, so anyone else who does must be wrong" really
__monty__ has quit [Quit: leaving]
iqubic has joined #nixos-chat
parsley936 has quit [Remote host closed the connection]
<bqv>
amazing, cause i use remote builders this one derivation just made a pointless 550MB round trip