<gchristensen>
(hint: duckduckgo for "rust substring")
<cole-h>
Also, something funny I noticed on https://nix.ci/: "...provides the CI integration..."
<cole-h>
gchristensen: lmao
<cole-h>
CI integration = ATM machine
<gchristensen>
hehe
<cole-h>
I didn't know you could hotfix that remotely :o
<cole-h>
Rather, without re-evaluating
<gchristensen>
hm?
<cole-h>
There is no more "tanti" -- it's the full `nix-instantiate...` stuff
<cole-h>
And (unless you did some magic backend stuff) it wasn't re-evaluted
<gchristensen>
oh, I deployed a fix and I guess it evaluated or something, I dunno, I didn't do anything
<cole-h>
:D
<julm>
looks like using prometheus is trending :) is it that good? I only "know" munin a bit
<gchristensen>
prometheus is pretty nice
<julm>
ack. I'll give it a try then
<gchristensen>
:)
<andi->
Yeah, prometheus <3 I have an i3status bar entry that shows me some of my stats that I care about (of my infra). Very nice since you can just query it via curl and filter through jq
<gchristensen>
:o need deets, andi
<andi->
I still don't have any unicode-character graphs but that is on my todo list :-)
<andi->
My current plan is to extend i3status-rs with a prometheus/grafana source since it alreayd has some code to render "graphs"
<gchristensen>
yeah or you could implement it all again in rust :)
<gchristensen>
looks like I'm upgrading my buildkite server to 20.03 sooner than later, the multi-agent support is too enticing =)
rardiol has quit [Ping timeout: 256 seconds]
<andi->
I kinda lost fun in writing rust code due to my attempts at using async in the pat 6 months.. So much wasted time for something that I'd have written with plain C and epoll in 10min
<gchristensen>
yeah, async is (still) a nightmare and I (still) avoid it
<andi->
It feels silly to spawn threads that block on sockets and then there is the whole story around interacting with plain sockets without some HTTP player around them... They might be used to send/receive and then the story around select is very weak because async....
<gchristensen>
yeah. I spawn a lot lot lot of threads.
<andi->
Then a few weeks ago I did run into some weak rust<->c bindings again... by the time I wrote a reproducer in C I had my entire program implemented in C... It will probably swing back in a few months. As long as I am the only user I don't bother to RIIR :-)
<gchristensen>
in that sense I am grateful I don't know C
<gchristensen>
different agents of mine have different privileges
<gchristensen>
it isn't wonderful, a job on one agent can escalate itself to other agents ....... but it isn't nothing
<julm>
heh, I know C, but not buildkite. I'm so old school :D
<julm>
thought "buildkite" was a hostname of a server of your's. an open-air refreshing one actually.
<gchristensen>
oh :)
<gchristensen>
all of my servers are named after minor futurama characters
<julm>
Good news everyone! that I know, I'm not that old! :P Well.. I know only major characters like Bender and Fry..
<gchristensen>
:D
<jackdk>
I name mine after Barkley, Shut up and Jam! Gaiden
<gchristensen>
nice
<drakonis>
perfect.
<drakonis>
julm: farnsworth?
<drakonis>
GOOD NEWS EVERYONE
nckx has quit [Ping timeout: 264 seconds]
nckx has joined #nixos-chat
nckx has quit [Client Quit]
nckx has joined #nixos-chat
nckx has quit [Remote host closed the connection]
nckx has joined #nixos-chat
<julm>
drakonis: **Professor** Farnsworth! right :D
nckx has quit [Remote host closed the connection]
<julm>
"Many episodes' major plot points are introduced by Farnsworth announcing, "Good news, everyone!"—either to unveil his latest invention or describe the company's latest delivery assignment, which is usually a suicide mission"
<gchristensen>
sounds about right
<julm>
On the very few occasions he has actual good news, he often opens with "Bad news, everyone!"
<danderson>
"Good news everyone! I've invented a device that makes you read this in my voice!"
<gchristensen>
hehehe
<gchristensen>
"that's not good news at all.."
<aleph->
Dangit
<aleph->
I should put Futurama on...
<julm>
The Professor rarely worries about the safety of the crew, viewing them as a means to an end, as evidenced in the first episode. After remarking that he was looking for a new crew for his intergalactic space ship, he was asked "What happened to your old crew?" His response was "Oh, those poor sons of... — but that's not important! What is important is that I need a new crew!"
<julm>
"Our crew is replaceable, your package isn't."
<gchristensen>
probably one of the better tv shows made
waleee-cl has quit [Quit: Connection closed for inactivity]
Jackneill has quit [Ping timeout: 250 seconds]
nckx has joined #nixos-chat
Jackneill has joined #nixos-chat
icey_ has joined #nixos-chat
iceypoi has quit [Ping timeout: 252 seconds]
rardiol has joined #nixos-chat
<etu>
Heya! :)
<gchristensen>
hey etu :)
<etu>
gchristensen: What's up?
<gchristensen>
me, apparently :P
<gchristensen>
how about you?
<etu>
Working, fixing things, looking at monitoring and stuff
<etu>
Migrating code from old framework to new framework
<gchristensen>
nice
<gchristensen>
public frameworks?
<etu>
Yeah
<gchristensen>
which ones?
<etu>
Silex... to Slim
<gchristensen>
nice
<etu>
They can be fairly similar, so I've spent a lot of time to make a good template last summer
<etu>
That behaves pretty much like most of our services does, with logging, error handling, configs, databases etc
<etu>
So now we're copy-pasting the template, fixing the controllers and routing and going through most of the code
<etu>
It's actually really nice. So much cleaning up :)
<etu>
And the bootstrapping is so much nicer now :)
<gchristensen>
very nice, that is great!
drakonis has quit [Quit: WeeChat 2.7.1]
avn has quit [Read error: Connection reset by peer]
avn has joined #nixos-chat
endformationage has quit [Ping timeout: 258 seconds]
<ldlework>
he's been working on that for as long as i remember
<Irenes[m]>
(the "neat" was re the palette thing)
<ashkitten>
Irenes[m]: i had to learn how to create riff palette files in order to get it working, lol
<Irenes[m]>
haha, fun
<ashkitten>
glimpse can export palettes in a number of human readable/writable formats, but pal files seem like the only easy thing to create that it can actually import
<ashkitten>
i will probably experiment with imagemagick or ffmpeg to automate the process more
<ashkitten>
the riff format is incredibly simple though, it's composed entirely of chunks that have a length and some extra data. the RIFF chunk just happens to be the entire file, and contains a subchunk of some type
<danderson>
the classic uber-fileformat: It can do anything! world: uses it for exactly one thing then continues inventing other formats
cole-h has quit [Ping timeout: 260 seconds]
__monty__ has joined #nixos-chat
<ashkitten>
lol, i found an off by one error in glimpse's riff pal loader
<ashkitten>
also it's extremely naive and uses hardcoded offsets
<ashkitten>
but i mean, it's a simple format so whatever
<ashkitten>
oh lol, it can also import css, which it is able to export
<srhb>
I tihnk assumptions might have just made an ass out of me. How do SPIDs/lwp (as reported by ps -T) relate to /proc/sys/kernel/pid_max? Does each lwp consume a pid? I might have to just create a program to check, because it seems the interwebs has conflicting information here...
lassulus has quit [Ping timeout: 258 seconds]
lassulus has joined #nixos-chat
<srhb>
My confusing is over this wording in man 5 proc: "PIDs greater than this value are not allocated; thus, the value in this file also acts as a system-wide limit on the total number of processes and threads." -- do they mean "and threads" as in _transitively_ you can have no more than pid_max * max-threads threads, or do they mean an lwp/SPID consumes a pid as relating to pid_max as well?
<srhb>
julm: Yeah, I've been all over the internet, and people seem to contradict each other.. man pthread_create helped a bit though, and I can reproduce failing pthread_create when I've created ~ pid_max threads.
<srhb>
So, TIL, I guess. :)
<julm>
"NPTL is a so-called 1×1 threads library, in that threads created by the user (via the pthread_create() library function) are in 1-1 correspondence with schedulable entities in the kernel (tasks, in the Linux case). This is the simplest possible threading implementation."
<joepie91>
cc gchristensen, I believe you were using it
<__monty__>
Are there any real E2EE conferencing programs?
<joepie91>
not any that can work at scale, that I know of, due to multiplexing requirements and bandwidth constraints
<__monty__>
That's what I figured : /
<__monty__>
Why does agreeing on a single symmetric key not work? That way there shouldn't be any more bandwidth required than non-encrypted solutions, right?
<__monty__>
Is it because it's impossible to authenticate people if you do?
<MichaelRaskin>
Scaling videochat to many senders usually means the server is combining the streams into one
<__monty__>
Sounds like a job for HOMOMORPHIC ENCRYPTION! (Picture superhero brandishing chest.)
<MichaelRaskin>
Superhero is, I guess, holding a waterhose because both servers and clients caught fire?
<MichaelRaskin>
I guess in _some_ cases, when one participant has a good desktop on a stable wired connection, one could deploy something open-source multiplexing there…
mrCyborg has joined #nixos-chat
<tilpner>
Do you really need to stream every video to everyone else? I don't video conferences often, but I think voice from everyone and one video stream should be enough?
<tilpner>
*don't do
<joepie91>
__monty__: single key encryption is a problem when participants join/leave, generally speaking in E2EE you're supposed to make content no longer accessible to people who have left the conversation
<joepie91>
so now you need key rotation, and that is quite difficult to orchestrate without a central party
<joepie91>
and you can't take the shortcut of separately re-encrypting things for each participant
<joepie91>
there are some proposed mechanisms to deal with this (separately encrypting a rotated stream encryption key etc.) but IIRC all of the proposed schemes have some edgecase or other security-wise
<joepie91>
anyway, time to tend to my garden :P\
<gchristensen>
yeah I know they're not
mrCyborg has quit [Ping timeout: 252 seconds]
Guest62967 has joined #nixos-chat
waleee-cl has joined #nixos-chat
<eyJhb>
Holy hell, Chromium has started to be really slow...
<etu>
Wasn't it always since several years?
<eyJhb>
No, it has been quite snappy until like 1 week ago
<__monty__>
Fake news!
<infinisil>
Wooo, stackoverflow (and co.) have a dark theme now!
<__monty__>
Hmm, has anyone else noticed the logos for FF and edge converging?
<infinisil>
I've been using a global dark theme before, but I didn't like it very well
<infinisil>
Probably because it's just mostly inverting the colors, which can often look weird
<joepie91>
infinisil: this one does a loooot better than that - it analyzes the page content and generates a page-specific dark theme
<joepie91>
it nearly always gets it right
<infinisil>
Instead I've been using Stylus to apply user-written CSS to specific sites, opt-in, which is working pretty well
<joepie91>
the tradeoff is that it makes things a bit slower to lead, especially huge pages, due to the analysis step
<infinisil>
joepie91: Ah neat
<joepie91>
load*
<joepie91>
(you can enable a static dark stylesheet for specific sites, and a few other modes, if the analysis mode somehow doesn't produce the desired result)
<joepie91>
it's genuinely a shockingly good extension, good enough that I think this is pretty much ready to be made a core browser feature
<joepie91>
which would also help with the performance thing :P
<__monty__>
Not free for safari :"( Guess I'll keep manually changing the css using the dev tools : /
<MichaelRaskin>
Not sure FF extensions are actually slower than built-in JS parts
<joepie91>
MichaelRaskin: the extensions aren't, but the DOM API is :P
<joepie91>
this kind of style analysis is pretty heavy on the DOM API
<MichaelRaskin>
joepie91: I have a suspicion that if such a thing would be integrated, it would still be JS
<MichaelRaskin>
And use X-Ray DOM
<joepie91>
MichaelRaskin: I doubt it - I think it would pretty quickly become clear that the performance is too bad
<MichaelRaskin>
By now it is pretty obvious that the idea of a browser as it is now is bad. And?
<joepie91>
this has nothing to do with browsers...
<MichaelRaskin>
Well, browsers that try to support both viewing HTML documents and running JS applications and doing both things weirdly and inefficiently by now — it is clear there are problems, it doesn't mean anyone is going to ever fix them
<MichaelRaskin>
I am not sure that built-in JS functionality even has better options than X-Ray DOM
<joepie91>
this seems completely unrelated to my comments regarding dark reader
<MichaelRaskin>
Erm? I said that integrated version will be basically the same thing as extension already does.
<joepie91>
I don't see why that would be the case. extensions are severely limited in how they can be implemented, code in the browser itself is not
<joepie91>
it leaves a lot more room for changing things around to do it more performantly
<joepie91>
eg. keeping metrics on styling during building/rendering of the page and using those for stylesheet generation rather than iterating through the DOM
<joepie91>
doesn't need to be, that's entirely dependent on the design
<MichaelRaskin>
More or less has to be, because this is a cross-cutting concern in a hot code path
<MichaelRaskin>
I think there was some pretty cheap trick to perform a post-rendering filter for the entire page, moving colours around
<joepie91>
it's not really, though; see above suggestion for example, the 'metrics collection' approach, that's a unidirectional data flow where the building/rendering process does not need to be aware of how the metrics are used
<joepie91>
the problem isn't in the actual color changing, the problem is in the analysis to figure out what the stylesheet should look like
<joepie91>
changing colors on a page is a solved problem, stylesheets are plenty fast enough
<__monty__>
The installer trick's pretty clever. And as long as apple doesn't try to prevent it calling it "malware" or "abuse" is pretty harsh.
<adisbladis>
apple--
<MichaelRaskin>
Well, it is clearly malware, and Apple just failed at securing.
<MichaelRaskin>
MS DOS 6.x had no security to speak of, it does not mean malware was impossible
<ashkitten>
heheheh every few months i get a new person in the github issues for my headset repo and it makes me feel like i did at least one worthwhile thing
<aleph->
Heh
<ashkitten>
i really need to clean up this code eventually and implement all the features from the library into the control program
<ashkitten>
i'm mostly just stuck on there not being a good way to work with structured data in a shell
<ashkitten>
what output format do i use so it's easy to parse?
<ashkitten>
etc etc
rardiol has quit [Read error: Connection reset by peer]
<aleph->
I mean depends on the shell, some work fine
<aleph->
pwshll, TmplOS terminal, etc :P
<ashkitten>
does... does templeos terminal use structured data?
<ashkitten>
but also neither of those were suggestions for a standard unix terminal
<gchristensen>
it gives you ring0 access and decompile / recompile code at any/all times ... it does have some cool structured data going on in some interesting places, including being able to embed art / diagrams right in code
rardiol has joined #nixos-chat
<gchristensen>
the videos of TempleOS are so cool
<gchristensen>
(except the bad ones, yikes)
das_j has quit [Quit: killed]
ajs124 has quit [Quit: killed]
<ashkitten>
oh
ajs124 has joined #nixos-chat
das_j has joined #nixos-chat
<ashkitten>
yeah i get scared seeing shit about templeos cuz a lot of it is just ableist people-watchy garbage
<gchristensen>
yeah, there is plenty of badness there
lovesegfault has joined #nixos-chat
<ashkitten>
ridiculous that this has to present treating mentally ill people with kindness as a novel idea, but i like what i'm seeing so far
cole-h has joined #nixos-chat
drakonis has joined #nixos-chat
abathur has quit [Ping timeout: 260 seconds]
__monty__ is now known as SirJection
SirJection is now known as __monty__
endformationage has joined #nixos-chat
<viric>
__monty__: I expect webrtc on filegive has conference end to end encryption
<viric>
that's two players
<viric>
But mesh can be done I hope - I haven't programmed any of that yet.
<viric>
poor man's videoconference.
<viric>
But I'm very annoyed at browsers not having working video encoding/decoding with vaapi or similar
<viric>
(on GNU/Linux)
<viric>
For what I've seen, 6 players in jitsi meet makes firefox (all cpu) do 4fps
<viric>
I'm surprised that noone cared on hw video decoding (not to say encoding) on firefox, for GNU/Linux
<viric>
I also built the chromium on nixpkgs 19.09 with vaapi and... apparently that doesn't work because I see the same cpu usage or more on video play. Yet about:gpu says Video Decoding Hardware Accelerated
<gchristensen>
the last part dealing with The Fire arrives tomorrow
<gchristensen>
replacing one of my hard disks which throws errors on scrub
<MichaelRaskin>
The GPU Fire?
<gchristensen>
yeah
<qyliss>
the gpu fire??
<gchristensen>
oh dear qyliss
<gchristensen>
you've missed so much
<joepie91>
to my amazement, it seems that I am going to get a next-day delivery from DHL Parcel(!) tomorrow
<joepie91>
for comparison, my boxes of Club Mate took a week to arrive, through the postal service, which normally does next-day stuff and is totally over capacity now
<gchristensen>
I accidentally caused a fire in my home server after plugging in a nvidia K80, and then the next day I bled on it trying to fix the problem with this disk throwing errors (I was thinking it was maybe an unseated cable) ... caused the system to no longer boot at all
<qyliss>
oh gosh
<joepie91>
while normally DHL parcel is the one known to screw up deliveries and get delayed...
<gchristensen>
it lit up like a lightbulb there for a minute
<danderson>
novel plasma-based rendering pipeline
<cransom>
literal blood sweat and tears.
<MichaelRaskin>
I dunno about you, but my lightbulbs do not smell of burnt dust
<danderson>
*may include minor side effects
<joepie91>
lol
<joepie91>
"duty cycle: not rated"
<ashkitten>
it works! it just doesnt work again
<gchristensen>
lol
<MichaelRaskin>
Reminds me of flashlight LED over-volting
<danderson>
way back in the day, google used to have computers on trays of rough cut metal. Sharp edges absolutely everywhere, nothing was safe
<gchristensen>
the good news is I didn't lose any data permamently, I only lost the 10 year old cpu/motherboard/psu that was in there, and one of my 3 drives in raidz1
<danderson>
hardware ops handed out tshirts and stickers that said "My other computer is made of razorblades and hate"
<gchristensen>
oh my word
<flokli>
gchristensen: I'll never ask you to do remote hands :-D
<danderson>
you got one if a rack made you bleed on the job
<joepie91>
and now Google just has sharp edges and hate in their software, rather than their hardware?
<joepie91>
:P
<gchristensen>
danderson: no wonder they took the approach of "jupst get another server [beacuse we're not touching that thing]"
<joepie91>
lol
<joepie91>
that does actually explain a few things...
<danderson>
heh
<danderson>
eventually there was a radical idea: the people designing the computers talked to the people who had to touch the computers
<gchristensen>
"mind rolling over those edges plz?"
<MichaelRaskin>
Also explains rumors of not touching racks until a fixed percent of servers die, then sending the entire rack to scrap metal
<danderson>
pretty much
<danderson>
"or just like any deburring at all, PLEASE"
<danderson>
yeah, that was always a bit of a legend-building exageration
<danderson>
hardware ops are constantly replacing and nursing single machines back to health
<danderson>
they just get to do it on a high-latency schedule
<gchristensen>
:)
<danderson>
high throughput, but a machine might sit dead in a rack for a week until they get to it
<gchristensen>
they would'nt want to lose too much blood in a day anyway
<joepie91>
danderson: honestly I didn't realize until now that the whole "every project is their own island and noone bothers to work with other people's expertise" thing at Google, also extended to the hardware...
<danderson>
lets them spatially optimize - pull 10 machines out of 10 racks near each other, take them to the repair station, etc.
<joepie91>
specifically that design
<gchristensen>
nice
<danderson>
instead of running 30 miles a day fixing individual boxes
<joepie91>
that makes sense
<danderson>
but google loves to exagerate the "lol yeah we just leave dead machines in the racks, we have sooooo many computers"
<joepie91>
I wonder if archive.org does something similar with HDDs
<gchristensen>
hehe I'm sure that is true
<danderson>
it was probably true at some point when google was okay burning money for 30% utilization
<danderson>
but then the CFO looked at the ledger and said you know, this is dumb"
<danderson>
similarly they love leaning on the legend of "commodity servers"
<danderson>
"yeah we just raid best buy and rack whatever lol"
<MichaelRaskin>
Hahahahaha
<adisbladis>
I always had the impression google hw was highly standardised
<MichaelRaskin>
(I have heard about overheat-tolerance-binning of CPUs)
<danderson>
today's google platforms are the biggest big iron you can get 2x0.5U form factor
<danderson>
96 core machines, 2x100G networking, ludicrous amounts of all the things...
<gchristensen>
once you get to a certain size density is a problem :)
<danderson>
it's still true that they're cheaper than buying that from Dell... But that's just because Google is its own OEM at this point
<MichaelRaskin>
If you get to a certain size and your problem is density, that means antitrust enforcement is asleep
<danderson>
I mean, it being america... yeah.
<cransom>
i was in an equinix while someone from google (before they were really a thing) was sitting on the floor, installing OS on machines in their cage from a thumb drive. it was trays of machines though, bare wire, no cases, zip ties.
<gchristensen>
well... yeah .... but I mean in terms of why have 50,000 servers taking up a lot of space when I could have 10,000 taking up much less
<adisbladis>
Heh, data center space is _expensive_
<danderson>
yeah, the machines still look very barebones
<cransom>
power/cooling density is usually the problem
<danderson>
but closer to Open Compute type compute sleds
<danderson>
(also means they don't rack the same kind of gear in colos - the rack form factor is different)
<nix-build>
#83896 (by etu, 5 minutes ago, open): PHP: Make the default package more sane [v3]
<danderson>
aah issue closure policies
<danderson>
aka passive-aggression as a service
<adisbladis>
gchristensen: nice
<gchristensen>
etu: !! nice!
<adisbladis>
I semi-accidentally took a few grams of caffeine once
<adisbladis>
It was absolutely horrific
<danderson>
pure caffeine has a surprisingly low LD50
<etu>
gchristensen: It became a co-lab project :)
<adisbladis>
danderson: ~180mg/kg body weight iirc ?
<MichaelRaskin>
danderson: I am not sure it is surprising that a strongly psychoactive substance has a low LD50…
<MichaelRaskin>
There is only so much excitement one's heart can take
<__monty__>
danderson: Hmm, I think I saw a picture once where they had machines on wooden boards rather than sheet metal. Sounds less razorblade-y.
<danderson>
__monty__: cork board, yeah
<danderson>
that's really old racks, before google started building their own computers
<danderson>
(pro tip: corkboard is flammable, which makes it a bad idea in datacenters... at least if you want your DC ops to like you)
<danderson>
but the cork board machines predate google building its own datacenters
<danderson>
the idea was: when you rent a cabinet in a colo space and get, say, a 20A breaker, they expect you to use maybe 10-15A of that.
<danderson>
google made sure that the contract didn't include any target load, and ran the consumption all the way up to just shy of tripping the breaker
<danderson>
which at the time you could only do by cramming computers in with no cases
<danderson>
but power density has improved since then, you no longer need to do that to max out a breaker :)
<joepie91>
that sounds like a good way to make your datacenter provider really dislike you
<danderson>
yeah, they were not fans.
<gchristensen>
_hopefully they had proper chillers, not fans_
<danderson>
lol funny you should say that...
<gchristensen>
oh no
<danderson>
there are photos of an early google colo rack with a desk fan sitting on the floor, pointed at the rack
<danderson>
but yeah. The colo providers got very cranky, but were locked into a contract. They expected to renegotiate mercilessly at renewal time, but then google just moved out and into its own datacenters
<joepie91>
this genuinely sounds worse than the infamous FDCServers cardboard box enclosures
<joepie91>
which is, well, I take my hat off
<danderson>
heh
<danderson>
the lesson IMO is you can get away with a _lot_ at small scale
<danderson>
and you don't have much to lose anyway :)
<joepie91>
sure, if you're willing to be an asshole :)
<MichaelRaskin>
It's actually a pity the DC operators did not find a way to break that setup without obvious liability
<danderson>
joepie91: which, early googlers... yeah.
<danderson>
there are some really nice people in that cohort, and some real pieces of work
<danderson>
ah well. The datacenter industry was still very young, and learned the hard way about being specific with contract terms
<danderson>
they're now really good at that :)
<gchristensen>
google was just finding bugs in the contract (corporate-apologist-level-9,000)
<MichaelRaskin>
Well, they might have been finding bugs in contracts with larger entity than themselves at least
<danderson>
and ruining it for everyone else who wants to buy colo space now
<MichaelRaskin>
They did go on to look for bugs in law enforcement
<MichaelRaskin>
(not even laws, as a ton of stuff they do has been ruled illegal more than once)
<danderson>
A thing I read that resonated with me a while ago: as the stakes get higher, laws turn into more like contract negotiation
<danderson>
the extreme example is country/country discussions, where there are no rules, you just figure out what you can get away with
<joepie91>
I mean, this is basically how "too big to fail" works
<danderson>
but it (sadly) maps well to large companies as well. It becomes less about "what is legal" and more about "what can we get away with, at what price"
<joepie91>
you just get yourself into a position that enough shit breaks in society when laws start getting enforced against you, and suddenly enforcement becomes a lot more optional
<joepie91>
right
<danderson>
it's pretty much directly referenced in google earnings calls these days. They budget out "we're going to pay this many billions in fines to the EU this year"
<danderson>
and it's just another line item on the financial report
<danderson>
and the universal reaction is "ah, excellent financial planning, good work"
<MichaelRaskin>
I wonder at what point EU starts using the fact that they created the laws allowing effectively unbounded fines from Google…
<danderson>
didn't they just do that? Or was it another company
<MichaelRaskin>
I think they did not cross 5B yet
<MichaelRaskin>
(per year)
<danderson>
I can't remember, but some company appealed a massive (hundreds of millions) GDPR fine in Ireland
<danderson>
and the judge's opinion was "you're right, the fine isn't big enough"
<MichaelRaskin>
Meh, 100M nobody cares
<danderson>
"no wait that's not..."
<MichaelRaskin>
I remember that the French 50M fine was interesting not because it's noticeable
<MichaelRaskin>
But because they basically said that Google Ireland is not really a company
<MichaelRaskin>
So Google is not represented in EU by an entity with any decision making powers
<MichaelRaskin>
So they won't respect single-window provision of GDPR
<danderson>
ah, I remembered correctly, it was google!
<danderson>
they appealed a $2.6B fine, and one of the judges said "you know, you're right, that doesn't seem high enough"
<danderson>
so, the legal strategy of "keep messing with the EU" is going great :P
<MichaelRaskin>
I mean, EU fines seem to be higher than total worldwide taxes
<MichaelRaskin>
But still bearable as taxes
<joepie91>
they've created plenty financial buffer for themselves through tax evasion, after all
<joepie91>
sorry, I mean "tax optimization"
<danderson>
yeah that's been the longstanding counter: "it's legal, if you don't like it change the law!"
<danderson>
not pictured: lobbying against changing the law
icey_ has quit [Ping timeout: 258 seconds]
<joepie91>
heh yeah exactly
<cole-h>
Anybody know of a (preferably single) git command that will show `tag-ghash` when ahead of a tag and `tag` when even with the tag? `git describe` also shows the number of commits ahead of the tag which I don't want/care about
<pie_[bnc]>
cole-h: git log or something similar had some kind of formatty thingy, but i dont know if it wil do what you want