<samueldr>
colemickens: nested virtualization is sometimes a premium
<colemickens>
samueldr: when you say premium, you mean resources? I figure and don't mind for what is mostly me messing around; gchristensen still digesting...
Jackneill has joined #nixos-chat
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-chat
<colemickens>
gchristensen: well, this is awesome (and more involved than I'd expected, I didn't realize the foxit repo just pre-builds/ships the clover/ESP image). Thanks!
<samueldr>
in my experience looking, nested virtualization is often only available on more expensive machines
<gchristensen>
you can get it at Amazon if you get the ".metal" variety, which are quite expensive
<gchristensen>
Packet of course
<gchristensen>
I don't know of many that you can get it from. maybe GCP has something
<gchristensen>
re clover: yes, and being so early in boot has fairly ultimate access, so you really want to be able to trust it if you want to be able to trust anything on that machine
<colemickens>
It's available for most all Azure machines, it was something they made easier than others early on, shockingly. GCP had (has?) some process where you have to modify the image itself, to enable nested virt
<gchristensen>
whaaaaaaaaaaaaaaaaaaaaaat
<gchristensen>
ugh, is this what finally persuades me to move beyond the goopiness of azure and use it?
<colemickens>
(I'm currently on the fly booting 96 core nixos instances to build stuff, they boot fast and build decently fast, push to cachix and throw em away)
<gchristensen>
where?
<gchristensen>
and: I can haz..?
<colemickens>
azure, westus2
<gchristensen>
is there a recent image, or do you make your own?
<Dotz0cat>
I tried asking in the nixos channel but I could not get an answer. can someone look at this package? I can get it to build but it does not display thumbnails for me. also since I am in a kvm I can not check if it plays any audio. https://gist.github.com/Dotz0cat/15a6e5e7ddb596605b6a8e0141c82c63
LnL has joined #nixos-chat
LnL has joined #nixos-chat
LnL has quit [Changing host]
LnL has joined #nixos-chat
LnL has joined #nixos-chat
LnL has quit [Changing host]
kalbasit has joined #nixos-chat
LnL has joined #nixos-chat
<cole-h>
Nice, I just ran out of space because of my ZFS snapshots on my win10 zvol... 👀
<cole-h>
Good thing it's real easy to delete snapshots for it
LnL has quit [Ping timeout: 256 seconds]
LnL has joined #nixos-chat
LnL has joined #nixos-chat
* colemickens
is unsure how people scripts aws, coming from azure. you can't name things like vpcs and subnets?
LnL- has joined #nixos-chat
<cole-h>
A question for ZFS folks: which snapshot service would you use if you wanted the snapshots to reside on another pool (basically: delete from "local" after sending to "remote")?
LnL has quit [Ping timeout: 264 seconds]
<cole-h>
e.g. I want to keep all my backups on an external disk to free up space on my local disk
<elvishjerricco>
cole-h: The majority of the zfs community has settled on sanoid/syncoid, but I find it rather unsatisfying. It's not nearly as good at keeping historical snapshots as some of the alternatives. I currently use znapzend, but I'm eyeing zrepl because it looks *way* more flexible and comprehensive
<ashkitten>
ughhhhh
<ashkitten>
this project is annoying me
<cole-h>
I currently use znapzend, but I don't think it has the ability to prune snapshots after they've been sent to the backup pool
<cole-h>
Or at least, not with the NixOS module
<ashkitten>
is anyone familiar with rust futures and could talk to me about how to redesign this library to use them
<cole-h>
I thought you wrote "features" for a second and was about to volunteer my novice eyes. However, I have no experience with futures :(
<cole-h>
I think LnL- has played with them some amount?
<elvishjerricco>
cole-h: My znapzend setup has a much shorter snapshot history on the local pool compared to the backup pool. Unfortunately I can't keep a more frequent but shorter history on the local pool; destinations have to receive snapshots equally as frequently.
<cole-h>
elvishjerricco: Oh, you can have different periods based on the destination?
<cole-h>
OH wait
<cole-h>
lol
<cole-h>
Missed the last few words
<cole-h>
Actually, missed the entire reply. Brain slow.
<elvishjerricco>
cole-h: Mine is set up such that I make a snapshot every hour, it gets sent to a backup server, and the backup server retains them longer than the local pool
<elvishjerricco>
zrepl lets you get way more complex
<elvishjerricco>
Like only sending new stuff to backups much less frequently; using holds to make sure it doesn't break its ability to keep going when the server is down
<elvishjerricco>
A horrible, horrible superset of json
<cole-h>
I only ask because I also hate YAML
<ashkitten>
grr my brain doesnt work
<cole-h>
Which zrepl config seems to use
<ar>
it's an example of a bad idea made even worse
<cole-h>
🤮
<elvishjerricco>
My favorite static config format is toml, and I don't even like that one all that much
<ar>
(json vs yaml)
<elvishjerricco>
I have found yaml implementations that don't parse all json for some reason. It's just so complex to parse that they can't even get the basic aspects of it down sometimes...
<elvishjerricco>
I don't even mean complicated json. I mean stuff like what builtins.toJSON would produce
<ar>
and just about every json implementation behaves differently in some cases
<elvishjerricco>
Yea isn't there some huge independent test suite that basically no json implementation passes completely?
<pie_>
i think me poking people is half the reason syncoid got merged
<pie_>
now youre telling me i should have found zrepl except its config is terrible :p
<ashkitten>
pie_: do you do rust
cjpbirkb1 has quit [Ping timeout: 258 seconds]
cole-h has quit [Ping timeout: 260 seconds]
Dotz0cat has quit [Ping timeout: 264 seconds]
<ashkitten>
i figured it out
<ashkitten>
how to the thing
<ashkitten>
be proud of me
* eyJhb
is very proud of ashkitten
<ashkitten>
yay
<ashkitten>
ty
<eyJhb>
What have you figured out?
<ashkitten>
uh
<ashkitten>
thing
<ashkitten>
with how to implement
<ashkitten>
a uh
<ashkitten>
thingy...
<ashkitten>
yeah sure ash
<ashkitten>
im very tired
<eyJhb>
10/10, do they thing ashkitten ! :p
<eyJhb>
the*
<etu>
,botsnack ashkitten
<{^_^}>
ashkitten: Oh thanks, have a cookie yourself
<ashkitten>
!
<pie_>
ashkitten: i wish i did rust
<elvishjerricco>
++ashkitten Figuring the thing out is always a rewarding accomplishment :)
leah2 has joined #nixos-chat
<elvishjerricco>
ashkitten++ ? Dunno how ++ works :P
<{^_^}>
ashkitten's karma got increased to 26
<ashkitten>
yay fake internet points
<ashkitten>
im hoping to get to 69 and keep it there forever
<elvishjerricco>
Lol is there a -- just in case you accidentally do something appreciated?
Dotz0cat has joined #nixos-chat
<eyJhb>
elvishjerricco: Yes, you ++ yourself :p
<eyJhb>
I have had negative karma for some time...
<{^_^}>
#96644 (by SuperSandro2000, 9 weeks ago, closed): rustscan: set platform to platforms.all #1234
<supersandro2000>
it does not react to its own messages
<supersandro2000>
a piets
<supersandro2000>
piety
<eyJhb>
We need another bot
<supersandro2000>
I once looped two bridges with each other
<supersandro2000>
noticed it after 2 Minutes aka over 1000 messages which got very long
<eyJhb>
samueldr: Do you have anything against OnePlus regarding firmware/porting/anything? Or are the not nice to work with?
<clever>
samueldr: i was recently looking into android auto, and ive found some interesting details
<clever>
samueldr: there are at least 2 opensource headunit implementations (the car/radio end), but ive not yet found a phone implementation (which would let mobile-nixos pop up on your radio)
<clever>
and the protocol involves openssl, where the headunit has its own certificate&key, to encrypt most comms
<clever>
the headunit can also return steering wheel angle, wheel speed, gps, gear, and info like is a passenger present
<eyJhb>
,ping
<{^_^}>
pong
aleph- has quit [Ping timeout: 260 seconds]
<clever>
eyJhb: and the pi foundation pulled a rabbit out of a hat with zero warning
<clever>
eyJhb: they released the pi400
<eyJhb>
pi400?
<clever>
the same general chip as the pi4, but now in a keyboard
<eyJhb>
Niiiiiiiiice
<clever>
the SoC also has a new revision (the 3rd i know of), which has more clock gating, to allow reducing the power usage more
<clever>
at a glance, its basically just a pi4b board, that has been stretched&squished a bit, and csi/dsi maybe be missing, 1 usb-host is also missing
<clever>
with an onboard keyboard controller of some kind
<eyJhb>
Maybe not the best time for this, for learners..
<eyJhb>
Sharing keyboards :p Unless they bring their own
<clever>
eyJhb: but i can see it being useful for learning@home, no longer worry about a naked pi board
<eyJhb>
True...
<eyJhb>
Speaking of stuff and things clever ! Have you seen OnePlus N10? :D
<clever>
nope
<eyJhb>
> DKK 2500
<{^_^}>
"8662525.000000 VND"
<eyJhb>
> VND 8662525
<{^_^}>
"329.175950 EUR"
<eyJhb>
For a phone that supports 5G :D
<makefu>
but what are the actual use cases for 5G? "Faster Internet" is certainly not a use case i have right now with 4g being reasonably fast
<clever>
> Scheduled to be killed in about 2 months, Google Cloud Print allowed users to 'print from anywhere;' to print from web, desktop, or mobile to any Google Cloud Print-connected printer. It was over 10 years old.
<nix-build>
error: syntax error, unexpected IN, expecting ')', at (string):345:24
<clever>
gchristensen: and in december, that will cease to function!
<gchristensen>
lol.
<hexa->
don't rely on google :3
disasm has quit [Read error: No route to host]
disasm has joined #nixos-chat
__monty__ has joined #nixos-chat
<eyJhb>
Why didn't people take the time, to write print(), instead of print in python?
<eyJhb>
I hate them.
<eyJhb>
Now I either have to patch the worlds simplest file, are use Python2...
<__monty__>
That was the only way to write it for decades though.
<sphalerite>
eyJhb: because it _wasn't_ a function :)
<sphalerite>
eyJhb: you can use 2to3 to patch it automatically though.
<eyJhb>
sphalerite: in 2019?
<eyJhb>
Because I am quite sure at that point, you have no excuses :p
<sphalerite>
ldlework: that looks right up your street haha
<ldlework>
hehe it doesn't look particularly useful, but it is cute
<joepie91>
I bet that'll be used in movies
<ldlework>
FWIW, it boots right up with appimage-run
<ldlework>
If you wanna try it
<eyJhb>
What..
<eyJhb>
,locate bin bzdiff
<nix-build>
Found in packages: bzip2.bin
<eyJhb>
:( RIP
andi- has quit [Remote host closed the connection]
andi- has joined #nixos-chat
<eyJhb>
Any ideas on building docker images, using Nix, but it is not feasible to convert all dockerfiles into NixOS Dockertools images?
<eyJhb>
Thinking of generating a shell script, that I can run afterwards to setup the whole env.
<viric>
my employer starting mitm-proxying all browser https connections by forcing employees to run a software that enforces new root CAs into the browsers....
<viric>
That looks very US or Chinese to me and I wondered how acceptable that is in EU. Has anybody any idea about this?
<eyJhb>
Not very acceptable as far as I know
<eyJhb>
Which country?
<viric>
the mitm is zscaler.com
<viric>
eyJhb: Catalonia under rule of Spain
<viric>
I see that the ECHR has a "Case of Barbulesco v. Romania" about this, that seems to matter on the domestic legislations
<viric>
zscaler.com has a whitepaper on GDPR-compliance as in all should be fine. The mitm-proxy is a local process that says to be doing the-right-thing.
<eyJhb>
Maybe someone knows better, I always tag joepie91
<eyJhb>
Also, I would reconsider the company
<joepie91>
viric: the exact details still vary by country I believe, but there is no general exception in the GDPR for employment scenarios; in fact, so far the legal opinion in NL seems to be that an employer literally *cannot* have consent from the employee, regardless of whether they ask, because of the power imbalance
<joepie91>
and so any data collection from employees must be justifiable on a basis other than consent
<eyJhb>
Also, on a pr. case basis I suspect ?
<eyJhb>
I love how knowledgeable joepie91 is.
<eyJhb>
joepie91++
<nix-build>
joepie91's karma got increased to -2147483648
<eyJhb>
<3
<joepie91>
:(
<joepie91>
viric: btw, is this on a company or personal device?
<eyJhb>
Doesn't it show the "real" score after?
<viric>
company device
<viric>
joepie91: the zscaler.com says that they don't do data collection; they run a local proxy in the device that does the https MITM
<viric>
(and uses Windows tricks to enforce the browsers to go through that proxy)
<joepie91>
data persistence is not a prerequisite for something to fall under the GDPR
<joepie91>
it applies to any sort of 'processing', whether that's persistent or not
<viric>
Basically all our webbrowser indication about TLS certificates is up to this proxy desires.
<joepie91>
viric: as for "local proxy in the device" - what is the purpose of the software?
<viric>
They whitelisted the bank websites, so that kind of indicates that they feel they do something wrong
<viric>
joepie91: detect malicious things comming from http, I guess
<viric>
+ ban access to websites not whitelisted
<joepie91>
is it actually 100% local?
<viric>
the proxy is 100% local, but it checks their cloud whether something is good or not. They say they "anonimize" anything they send to their cloud
<viric>
I tried a twitter login (before I learnt this was there) and twitter asked me whether I was a robot for the first time.
<joepie91>
yeah I don't buy that :)
<supersandro2000>
so if it is local I can bypass it
<joepie91>
viric: does it send *every* unknown thing to a server, or just things that are flagged as probably-malicious?
<viric>
I can't tell
<viric>
it's blackbox software
<joepie91>
viric: how big is the company? (proxying questions from someone else now)
<viric>
supersandro2000: if you tell the browser to distrust its CA, it somehow rules the browser again to trust it. It deals with the browser.
<viric>
joepie91: In nasdaq.
nix-build has quit [Remote host closed the connection]
<viric>
The policy comes from the US, so I wonder how they can easily apply that to the EU. The zscaler.com own whitepaper on European GDPR claims that they are good, I guess.
{^_^} has joined #nixos-chat
<joepie91>
viric: "then a DPIA should have been carried out"
<viric>
supersandro2000: the browser doesn't look configured to use that proxy but it uses that proxy.
<abathur>
I saw sbang on lobsters and just knew I'd find gchristensen in it somewhere :)
<gchristensen>
hehehe
<viric>
joepie91: is that a document employees should have been given?
<supersandro2000>
It just if you have local access to a device you can always just wipe it and install whatever you want
<supersandro2000>
such local things only prevent non techies from doing whatever they want
tilpner has quit [Quit: tilpner]
<joepie91>
viric: unsure. but you could ask the DPO in the organization
<joepie91>
who should know about this
<joepie91>
the DPIA is basically the assessment that justifies the necessity of measures like these
<viric>
joepie91: so a document that we should be able to read, kind of.
<viric>
supersandro2000: I suppose they enforce the running of this software. Not optional.
<viric>
supersandro2000: and I don't mean technical enforcing to cheat about.
cole-h has joined #nixos-chat
<supersandro2000>
if you invest enough time you can bypass it
<etu>
Maybe one should start using flakes for the config
<etu>
🤔
<joepie91>
viric: one thing you should certainly be able to do, is request exact logs of all your data that has been processed
tilpner has joined #nixos-chat
<viric>
that's easy to say
<viric>
maybe they don't log to avoid persisting data :)
FRidh has joined #nixos-chat
<eyJhb>
etu: Sigh...
<eyJhb>
I just get this thing running?!
<eyJhb>
got*
aleph- has joined #nixos-chat
LnL- is now known as LnL
<supersandro2000>
Did you ever try to buy something from Steam an hour before the sale ends? No, let me tell you about my journey:
<supersandro2000>
First the shop 502'd with their referal code at the end
<supersandro2000>
then paypal did not work and timeout
<supersandro2000>
and finally they did not accept the special character in my bank login
<supersandro2000>
so spending money is a journey now days
* cole-h
hates PayPal
<cole-h>
I tried to use them to pay for something once, but they wanted a photocopy of my SSN lol
<supersandro2000>
a friend of mine once tried to buy something at apple and his credit card got locked
<supersandro2000>
nice. they booked the money but I did not receive my putchase
<supersandro2000>
Steam Customer Experience while they have sale: 502 nginx
<joepie91>
supersandro2000: other people are getting 502s also
<joepie91>
looks like they're just bork
<supersandro2000>
yeah, sadly. I experience this since at least 5 years
<supersandro2000>
I don't know how they still can't fix this
<supersandro2000>
it is every sale that it isn't funny anymore
<samueldr>
[03:45:31] <eyJhb> samueldr: Do you have anything against OnePlus regarding firmware/porting/anything? Or are the not nice to work with?
<samueldr>
recent devices have a timed bootloader unlock AFAIK
<samueldr>
otherwise they release(d) sources promptly
<samueldr>
recent is either starting with the 6 or 7
<samueldr>
other than that, be warned that SoCs in the snapdragon 845 vintage and more recent may not allow you to use graphics as currently implemented for the time being
<samueldr>
as vendor tree from qualcomm has broken the fbdev driver
<samueldr>
I need to, at some point in the future, make a DRM-based renderer for the parts made for Mobile NixOS
rajivr has quit [Quit: Connection closed for inactivity]
FRidh has quit [Quit: Konversation terminated!]
<colemickens>
can you just boot some small gui app in Cage?
<colemickens>
I have to `pulseaudio -k` every time I boot to get my usb audio device to show up. hm
<samueldr>
colemickens: you tell me :)
<samueldr>
colemickens: you have the hw and sofware setup to test it
<samueldr>
(and the knowledge about wayland I don't)
<samueldr>
but for stage-1, I don't know if that's a good way to go, vs. simply adding a branch for DRM rendering
<samueldr>
since AFAIUI you can get a framebuffer that works just the same way
<samueldr>
so it's mostly initialization stuff that's needed
<colemickens>
I see
<samueldr>
everything else (on razer-cheryl2) has been verified to work
<colemickens>
samueldr: I find it hard to want to work on it without knowing if WiFi will work, and so then my thinking is that I'd rather focus on booting mainline, I don't know if that's misguided.
<samueldr>
not misguided
<samueldr>
it's totally understandable to want to have some kind of external communication going
<samueldr>
and the fact that SDM845 has some mainline work ongoing makes it a good target here for mainline
<colemickens>
well, I don't actually have data to prove that mainline is more likely to have wifi working.
<colemickens>
it's mostly that, yeah, the other support for SDM845, that made it seem worth pursuing.
* colemickens
also day dreams about waiting a year or two for GKI devices to be more widely available...
* colemickens
actually might try booting an Android 11 GSI on Pixel 3 for fun
<samueldr>
I've read that for mainline and SDM845 wifi is expected to work
<samueldr>
I don't know that GKI will help in any way with non-Android systems
<samueldr>
it could even cause harm by having vendors not share source code of their modules through contorted wrong logic
<samueldr>
>> BuT GoOgLe gAvE yOu ThE SoUrCe oF tHe KeRnEl!
kalbasit has quit [Remote host closed the connection]
kalbasit has joined #nixos-chat
<ashkitten>
pipewire pulse-server....
<ashkitten>
excite
<ashkitten>
might be able to finally actually use pipewire full-time!
<colemickens>
Oh right, I was aware of that being an issue for fuschia but hadn't considered that for GKI. I kind of assumed that maybe the community would be able to keep porting the GKI driver interface so we could re-use vendor drivers, but I realize that probably won't work out that way, and isn't what we want in terms of freedom. :((
<samueldr>
the only real solution to this problem is the way the chromeos team does it: always send patches upstream
<samueldr>
that's not a solution that solves every problem; you'll see it in e.g. gru scarlet, where their driver is mainline, but is buggy because it doesn't implement some things properly
<samueldr>
but at the very least the code is present
<ashkitten>
what are yalls favorite music players?
<samueldr>
rockbox
<ashkitten>
i meant on linux lol
<samueldr>
ah
<samueldr>
I have none on Linux
<ashkitten>
ah
<samueldr>
since all seem to be not nice compared to my top two
<samueldr>
rockbox and old-windows-only musikcube
<samueldr>
what I liked about that older musikcube (when I was using windows) was how its dynamic playlists were made
<samueldr>
you could make an SQL query that e.g. randomized and ordered by least played so you would go through your library instead of "actually random" which sometimes ends up playing the same tracks over and over
<samueldr>
rockbox is just the bee's knees, sad my iPod seems to be having troubles (I haven't looked into it yet)
<samueldr>
currently using cmus
<ashkitten>
the closest thing to the nice spotify ui i've found is lollypop
<ashkitten>
but lollypop is weird
<ashkitten>
it makes very strange ui decisions
<ashkitten>
how's elisa?
<ashkitten>
apparently "not working"
nullheroes has joined #nixos-chat
V is now known as ^
^ is now known as V
<gchristensen>
trying to write a nix test, having to recompile nix every time I change the nixos test framework test because the source of the framework test is part of the source of the nix package
endformationage has joined #nixos-chat
cjpbirkbeck has joined #nixos-chat
cole-h has quit [Quit: Goodbye]
* colemickens
regrets some of the not nice things he said about azure after using aws
<gchristensen>
lol
<gchristensen>
devil you know
<colemickens>
for sure
<gchristensen>
aws was nicer before it had so many services
<gchristensen>
and its console was less powerful for niche stuff but easier to use. doing advanced stuff -> had to use the api. now you're punished by paying the complexity cost constantly in the UI
<colemickens>
Azure I think also benefited from getting to do a v2 api when transitioning classic to ARM.
cole-h has joined #nixos-chat
<danderson>
it's also a function of how AWS operates. Each product is a mostly independent company, in effect. And there doesn't seem to be any group in charge of making the console as a whole not insane.
<danderson>
so, it's really a hundred or so different companies, each building one product, all crammed into the same UI.
<colemickens>
afaict there's still no alternative to azure's resource groups, no built in way to delete resources by tags, and awsweeper uses terraform under the covers. hm.
* colemickens
should just boot a packet instance and call it good
<gchristensen>
aws-nuke is a thing iirc but I couldn't package it for osme reason
<colemickens>
hm, Equinix Equi-nix, is it a conspiracy?
<cole-h>
👀
<colemickens>
gchristensen: thanks; I'll peek at that too
<danderson>
yeah, azure resource groups are pretty amazing, I'm confused why no other cloud stole that
<joepie91>
danderson: do you have a 5-minute "about to go to bed" cliff's notes version of what they are?
<danderson>
they're a collection of cloud resources, that you can manage as a unit
<danderson>
think VMs, role accounts, storage buckets, managed DBs, etc.
<danderson>
it's just a unit of management of your cloud stuff. Instead of having a pile of random VMs and stuff, you have one resource group for "service A" that contains everything that service needs, one resource group for "service B", etc.
<danderson>
makes it much easier to keep track of why stuff exists (just look at the resource group that owns it), and to spin up/tear down ephemeral things cleanly
<gchristensen>
nice!
<gchristensen>
I can't believe AWS doesn't have that.
<gchristensen>
could you do a resource group per CI run?
<samueldr>
what if they A/B tested it and saw profits reduce from accidental usage?
<samueldr>
might be too cynical
<danderson>
I suspect it's more that it would require all products to adopt a company-wide thing
<danderson>
which is just now how AWS functions organizationally
<danderson>
not*
<gchristensen>
I would, if I could, create ephemeral AWS accounts just to bodge in support
<colemickens>
... and the only way to tag aws resources at creation time is to respecify the tags, and the type, for every possible resource type that might be created.
<colemickens>
I am just... Idk
<danderson>
yup, AWS is a tire fire for resource management
<danderson>
and that's basically the entire reason Terraform exists.
<danderson>
the invisible hand of the market fixed the glitch, by having another company make the missing product
<colemickens>
gchristensen: Azure actually went "up" and added multi-level subscription management for further grouping (so you can create harder subscription boundaries but still sharing billing params)
<gchristensen>
nice
<colemickens>
I was kind of wondering if indeed AWS's answer was some sort of ephemeral account but seems not. Seems like the community tooling is important
<gchristensen>
plz don't make me like azure
<gchristensen>
(I'm kidding)
<samueldr>
I only compared azure's pricing against a dedicated server for a specific use case a past client had... long story short it would have cost more to save a few bucks in their case (no devops kind of employees)
<samueldr>
how's azure compared to other clouds?
<samueldr>
(it would have cost more as the client wouldn't have been able to use cloud native stuff, and it would just have been a dumb reserved instance instead)
<colemickens>
My guess would be that the comparison to a dedicated server would be about the same.
<samueldr>
I meant, is azure generally in the same price range than google's and AWS are?
<danderson>
yeah, they're all comparable
<danderson>
price-wise, that is
<danderson>
Azure has much better Windows and Active Directory support, unsurprisingly.
<samueldr>
I know full well that for the "just a server" use case it's not even an option
<colemickens>
I can't believe gcp/az lack aarch64 still. :/
<danderson>
Azure also had some early issues with network reliability, and current issues with supply vs. demand
<gchristensen>
I'm not surprised, aarch64 is super annoying
<joepie91>
danderson: right, so it's basically a categorization thing?
<samueldr>
I believe amazon's AArch64 support is mostly a vertical integration thing for them
<danderson>
covid greatly increased cloud demand for some reason, and azure was caught short.
<joepie91>
danderson: or is there more to it
<samueldr>
if they could have made their own x86_64 hw they probably would have
<danderson>
AWS's secret weapon is that they're really, really good at logistics, and can do JIT capacity deployment at remarkable speed
* colemickens
hmms
<danderson>
GCP's been trying to learn about that and catch up.
<danderson>
from a former capacity manager at AWS, I learned that AWS effectively stockpiles computer parts in warehouses near the datacenter, and does final assembly at the last mile. Which lets them go "well this week, we need more of server shape X, so we'll build some of those"
<gchristensen>
whoa
<gchristensen>
nice
<samueldr>
wouldn't rectangular prisms be more efficients than X?
neeasade has quit [Remote host closed the connection]
<danderson>
contrast this to more traditional datacenter build-out, where you place an order for a couple hundred machines, and your assembler in Taiwan says "sure, they'll be there in 9 weeks"
<danderson>
that makes AWS amazingly good at responding to shifting demand, and as a result they basically never run out of computers to sell you.
<danderson>
contrast to GCP and Azure, which afaik use more traditional procurement methods, and rely on correctly guessing future demand months ahead of time
<danderson>
when it works, it's great. When they mispredict... "sorry, this region is out of capacity, try again later"
<gchristensen>
isn't amazon a book store
<samueldr>
pivot!
__monty__ has quit [Quit: leaving]
<gchristensen>
the 60 minutes documentary (about 1h+) about amazon was very interesting until it got weird
<danderson>
that's amazon's amazing secret: they're actually a logistics company.
<danderson>
that's their main core business, and they're very, very good at it.
<danderson>
but by day, they pretend to be a shop and a cloud
<cole-h>
Oh man
<cole-h>
I was feeling kinda meh today, but getting praised on my writing has uplifted my mood like 10 steps.
<cole-h>
(Context: turned in a paper last Sunday that I was hoping to get at least a B on -- after class today, prof submits the grade and it's a 104!)
<infinisil>
Ughh, I should never have opened the windows
<infinisil>
All the farmers around our house sprayed the grass with dung..
<infinisil>
Now my whole room smells like cow poop
<V>
ow
fnlaai has joined #nixos-chat
<fnlaai>
hi there
<fnlaai>
what's up?
<samueldr>
infinisil: give it a good rest and you won't smell it, only smell of it
<fnlaai>
hi samueldr what are you both talking about?
<infinisil>
Yaaay
<infinisil>
(logs are in topic)
<samueldr>
follow-up to a discussion that started before you joined
<fnlaai>
ah, my bad, I already checked out, it was cow poop
<infinisil>
🐄
<fnlaai>
How is daily life in Switzerland under pandemic infinisil?
<infinisil>
For me pretty chill, as I pretty much don't need to go out at all :)
<infinisil>
Today I went grocery shopping, and everybody wore a mask! I think it's mandatory, but still!
fnlaai has quit [Ping timeout: 245 seconds]
* colemickens
rigged up an `awsweeper-tag` that just deletes all resources with a certain tag, and then use `aws ec2 create-tags` to bulk add a tag to all the resources I create in the deploy script
<samueldr>
you know how there's street smarts?
<samueldr>
I think there's cloud smarts
<samueldr>
and that's why I don't do cloudputering
<samueldr>
I'm not a cloud-smart at all
<samueldr>
way too much to learn about to feel safe using, not enough time