<{^_^}>
prometheus/node_exporter#1050 (by grahamc, 6 minutes ago, open): smartmon.sh: Shellcheck, and support more OSes
<__monty__>
Can you tell shellcheck not to complain about a missing .sh extension? We have a polyglot and having to rename everytime you want to shellcheck is annoying.
<gchristensen>
it doesn't care about .sh
<gchristensen>
so something else must be wrong
jD91mZM2 has joined #nixos-chat
mmercier has quit [Ping timeout: 265 seconds]
lopsided98 has quit [Ping timeout: 260 seconds]
lopsided98 has joined #nixos-chat
<__monty__>
gchristensen: Ah, it wasn't the extension. But it refuses to check our file because it has a python shebang.
<gchristensen>
you can do --shell=bash I think?
<__monty__>
It's designed to be run as python or sourced as bash.
<__monty__>
Specifying the shell it still complains about the shebang.
<gchristensen>
hrm, this doesn't emit any warnings: shellcheck -e SC1071 --shell=bash projects/nixops/coverage-tests.py
<__monty__>
Indeed it doesn't.
<__monty__>
For my file either.
<__monty__>
I think I did get warnings when I extracted the shell from the polyglot.
<__monty__>
I'm not sure it's useful though because I don't know if the semantics of running and sourcing shell are the same.
<gchristensen>
largely they are
<__monty__>
So following the warnings for executed shell won't mess with the current semantics of sourcing it? I'm not familiar enough with shell to confidently do this. And applying the "improvements" blindly might lead to headaches, since I don't understand it in depth now and I'd have to make a bunch of changes that I have no idea what their individual effects will be.
qyliss^work has joined #nixos-chat
* joepie91
waves at qyliss^work
<qyliss^work>
hey
<joepie91>
qyliss^work: to clarify, I'm arguing this from the perspective of having seen various projects go 'open-core', almost invariably followed by sketchy things occurring, and OSS usage essentially becoming a second-class citizen; the switch to an open-core model says a lot about the intentions of the developers and how/who they intend to support... but the difference between open-source and open-core is too subtle for many people to really *see* the
<joepie91>
difference, leading to the prevention of forks
<joepie91>
so you kind of get stuck in the situation, as an ecosystem, that things are bad enough to be problematic; but not bad enough to motivate people to fix it
<qyliss^work>
Right, that makes sense.
<joepie91>
I've pretty much come to consider open-core models as a death knell for OSS projects
<joepie91>
the first signal to start moving away from them
<qyliss^work>
I just think that in this case, the issue has so much publicity that people will be aware of the difference
<joepie91>
it's also why I don't consider them OSS anymore from a pragmatic perspective; the idea behind OSS is a shared commons of software [that can support other software in that commons], and when a project is run in such a way that it's effectively useless to the commons...
<joepie91>
qyliss^work: have you heard of influxdb?
<qyliss^work>
only very vaguely
<joepie91>
well there you go :)
<joepie91>
influx went through a similar "oh noes" cycle
<joepie91>
very little knowledge of it lasted beyond the initial outrage
<joepie91>
the angry hackernews comments and whatnot can make it *look* like there's widespread knowledge of the problem... but in reality it's not likely that more than a handful of people are aware of the issues in the long term
<joepie91>
hell, I suspect, many of the angry commenters don't even understand the situation *right now*
<qyliss^work>
yeah
<joepie91>
what frustrates me immensely is that a few years ago I ran into a *really* well-written article that summed up the conflicts of interest and subtle consequences of different OSS funding models
<joepie91>
but it vanished from my bookmarks somehow, and I've been unable to find it again since
<joepie91>
it also covered a lot of things like how offering commercial support tends to lead to worse documentation since there's no incentive (even a disincentive!) to fix it, things like that
<joepie91>
and I feel like there's a really poor understanding in the OSS community of the effects of funding, so that article is really needed :?
<joepie91>
:/*
<qyliss^work>
that point on commercial support / documentation is a good one
<qyliss^work>
yeah, that sounds like a really good article
<joepie91>
the closest thing I know of today is https://github.com/nayafia/lemonade-stand but it doesn't quite put it into a coherent overarching narrative, and iirc last I looked at it it was missing a few things
<joepie91>
qyliss^work: the depressing conclusion is that basically every funding model has significant tradeoffs / conflicts of interest, with donations being the least impactful but also much more work to get right
<qyliss^work>
yeah
<joepie91>
(donations can be zero-impact depending on situation, but aren't always)
<joepie91>
from a more technical perspective, one major issue with OSS funding is that there's basically no donation infrastructure that isn't either extortionate or limited or both
<joepie91>
like, I can't say "I have a great model for handling donations and distribution, I'm going to build a service to do it!"
<joepie91>
because now I'm either a money transmitter or have to deal with an API that's very poorly designed for donations
<joepie91>
(or that charges outrageous fees)
Lisanna has quit [Quit: Lisanna]
<qyliss^work>
What's Patreon's fee structure these days?
<qyliss^work>
Seen a few projects using that
Lisanna_ has joined #nixos-chat
<joepie91>
qyliss^work: briefly bad and then un-bad after people were outraged; but Patreon is very limited
<joepie91>
it doesn't allow you to design your own donation system
<gchristensen>
patreon has roughly 10% overhead
<qyliss^work>
oh wow
__Sander__ has quit [Quit: Konversation terminated!]
<gchristensen>
it isn't so bad, and not outrageous for that sort of platform
<__monty__>
I hope something like bugmark catches on.
<joepie91>
'market-driven' sounds like a great way to make things worse :)
<joepie91>
also it has a blockchain for some reason?
<__monty__>
It's a good use of smart contracts imo.
<__monty__>
It being on a blockchain gets you the advantages of decentralisation for free.
pie_ has quit [Ping timeout: 264 seconds]
<__monty__>
joepie91: I actually think it could be a big boost to open source. There's people that are knowledgeable enough to contribute but can't spare the time because they need to do paid work instead. If contributing to open source turns into paid work, a bunch of people will be able to do open source instead of proprietary work.
<gchristensen>
I want to live in that utopia
<joepie91>
__monty__: thing is, the problem you're describing is *caused* by software development being market-driven; you don't solve a problem by applying more of the problem...
<joepie91>
this seems to be yet another project that's based on the faulty premise that if it's valuable, 'the market' is willing to pay for it
<joepie91>
the correlation is much vaguer in reality, and the mismatch between the two is why OSS falls between the cracks
<joepie91>
in particular for things like security issues and review, which seems to be one of the things targeted by that project specifically
<__monty__>
No, current markets are not the problem. The problem is they're not accessible enough. 1) Positions to get paid for open source work are rare, so devs don't have access. 2) It's too hard to get enough funds together to properly finance development as a user. A decentralized solution allows many people to cooperate to sponsor development.
<gchristensen>
never have I ever met a company with money to spend say hooray let's pay for a thing we can get for free, and raises for everybody
<joepie91>
__monty__: people can _already_ cooperate to sponsor development, that isn't the problem
<__monty__>
There's no decentralized solutions for it though.
<gchristensen>
where is decentralization the problem?
<__monty__>
Bug bounty sites create the potential for a new facebook that shows devs the projects that pay for exposure.
<joepie91>
how is decentralization relevant?
<gchristensen>
it takes me 45s to mail you a cheque for some work you've done via my bank's ui, I just don't want to
<__monty__>
The bounties would be targeted, just like ads are now.
<__monty__>
gchristensen: There's plenty of people that want certain features, some of those people are willing to contribute towards paying for them.
<joepie91>
and they already can?
<__monty__>
joepie91: And they already do, what's your point?
<__monty__>
Decentralization prevents the horror scenario I described above.
<joepie91>
my point is that you've failed to support the claims that 1) decentralization is necessary to solve a problem, and 2) the problem is that people don't have a good means to pay
<joepie91>
__monty__: what horror scenario?
<joepie91>
and how does it prevent it?
<gchristensen>
this is hardly a "build it and they'll come" scenario
<__monty__>
gchristensen: I'm not so sure. I'd use bugmark. I don't care for existing bounty sites.
<gchristensen>
I've found Patreon to be an effective way to pay for the ofborg infrastructure
<__monty__>
The problem with the patreon model is patreon gets rich not doing much of anything, patreon gets to censor who gets money and don't you have to pay regardless of whether work's being performed?
<gchristensen>
they do loads of things
<gchristensen>
they have made it so simple to do, I am really quite happy with the service and cost
<joepie91>
__monty__: of all the actual problems with OSS funding, that one is pretty low on the list :)
<joepie91>
it's not a nonzero problem, to be clear
<joepie91>
but it is not The OSS Funding Problem
<andi->
there was an alternative to patreon that many people switched over to that doesn't have a markup.. When I signed up to patreon it felt a bit silly to give them money just for passing it along..
<__monty__>
I won't ever contribute via patreon. Which is too bad because that means it's hard or even impossible for me to contribute to several projects.
<gchristensen>
__monty__: I can send you my address if you want to mail me a cheque
<andi->
I can tell you first hand why companies potentially aren't doing more OSS. My current employer has like 500k users (on thousands of instances) on our open source version. We have a few hundred paying customers at best... Now investors really don't feel like it is paying off for them..
<__monty__>
Centralization is not a problem to be trivialized.
<gchristensen>
__monty__: I can send you my address if you want to deliver gold to my home somehow
<joepie91>
errrr
<__monty__>
gchristensen: None of those is convenient enough.
<joepie91>
it's a nonzero problem*
<joepie91>
sorry
<joepie91>
unintentional double negative there
<__monty__>
I also lose any guarantees you actually solve the problem I want you to or you lose the guarantee of getting my money if we reverse the order.
<gchristensen>
it is true
<joepie91>
__monty__: this is generally true for any kind of thing that is not provable automatically without any third-party input
<gchristensen>
andi-: fwiw I'm really really happy with Patreon, and find it provides more value than they charge... so please don't feel bad about it on my behalf
<joepie91>
(yes, "prediction markets", I am well aware of them, I've seen precisely zero proposals that are resistant to serious adversaries)
<joepie91>
andi-: honestly, part of the problem there is seeing OSS as a business consideration
<andi->
joepie91: well the current world evolves around money. I rather work somewhere where I can do a bit of (F)OSS an not just internal work when I have the chance. Sadly I also still need money to buy food :/ But yes I agree..
<joepie91>
andi-: sure, and I get that, but I think it's a mistake to try and sell OSS to a company as something that benefits the company; it just has the inevitable effect of said company trying to distort the concept to a point where it's not really OSS anymore, but it looks close enough and makes them money
<joepie91>
the point of OSS is a public commons; that is a valuable goal in and of itself, business considerations notwithstanding
<joepie91>
and in practice, if you have a full-time job, this is pretty much something you need to negotiate as an employee
<joepie91>
(eg. "work must be at least X% third-party OSS work", or even "all work must be OSS")
<andi->
I do have that.. I am allowed to work on OSS during work hours - in a limited amount.
<joepie91>
this is a lot easier if you're a freelancer :P
<andi->
But you asked why not more companies would consider OSS a viable thing. Most business people do not care about public commons..
<andi->
they just care about their new car, house, pay check, ..
<__monty__>
joepie91: At least with a smart contract both the user and the developer have a guarantee that if the project maintainer isn't malignant that they'll get what they paid for, get paid for what they did. Yes it still depends on the project maintainer BUT it gives more than no guarantees whatsoever.
<andi->
and they gladly use OSS if it saves them money..
<gchristensen>
what if there is a vulnerability in the smart contract and an attacker steals the funding?
<andi->
the computer is perfect!!1 ;0
<joepie91>
__monty__: I don't see what this meaningfully changes. either the project owner is honest and it doesn't change anything compared to not having a smart contract, or the project owner is dishonest and they subvert the contract and it *still* doesn't change anything compared to not having a smart contract
<gchristensen>
because I know what happens if a contract of mine fails to pay for work done
<__monty__>
gchristensen: There's literally no solution that gets around that. And I'm not just talking about technical solutions.
<joepie91>
[19:05] <andi-> But you asked why not more companies would consider OSS a viable thing. Most business people do not care about public commons..
<joepie91>
huh?
<joepie91>
are you sure you're not thinking of a question somebody else asked?
<gchristensen>
I mean, there is a solution, because I've done it, small claims court has worked great for me in the past
<__monty__>
joepie91: Well, the maintainer never gets his hands on the money. Money corrupts. In this case the maintainer can either accept the work, helping him improving his project, or refuse the contribution, effectively shooting himself in the foot.
<andi->
joepie91: probably, not re-reading the backlog now..
<__monty__>
gchristensen: Your small claims court probably works in most cases. I'm not so sure it'll keep working against future megacorps.
<joepie91>
__monty__: they can just as easily claim to have done the work without actually having done it
<gchristensen>
(it does)
<joepie91>
this changes nothing
<ivan>
andi-: those numbers are wild. I'm surprised they've held out that long.
<joepie91>
honestly this argument really feels like rearranging the deck chairs on the titanic to me :)
<gchristensen>
yea
<andi->
ivan: well the product isn't the main income or wasn't until recently ;)
<__monty__>
joepie91: Ok so the maintainer claims the work was done. They end up with no contribution, because the work wasn't done and the dev gets paid for doing no work. What's the maintainer's motivation?
<__monty__>
I'm not saying bugmark is already *the* solution. It's definitely an improvement. I'd like to eliminate the need for a trusted maintainer but it's not an easy problem.
<joepie91>
I mean, if you add a trusted party into the mix, sure, it's easy to solve
<gchristensen>
hold the phone
<joepie91>
but then it is also already solved; this is what escrow is for
<joepie91>
and has been solved for decades
<gchristensen>
oh man
<joepie91>
and that basically delegates bugmark down to "like escrow agents, but on the blockchain", at which point... yeah
<__monty__>
No, it's not the same. Without the smart contract. You'd have the maintainer escrow the money. That gives him a great motivation to claim the issue was fixed, he can keep the money.
<__monty__>
With the smart contract he doesn't have that motivation.
<gchristensen>
are there any other interesting things to discuss?
<gchristensen>
so, wait, how does the smart contract change the motivation?
<__monty__>
gchristensen: The maintainer can only push the buttons execute the transaction or cancel the transaction. In neither case they get the money.
<gchristensen>
who is the maintainer?
<__monty__>
Whoever maintains the open source project.
<joepie91>
¯\_(ツ)_/¯ the maintainer can just create a fake developer and pocket the money
<gchristensen>
and how does a smart contract add this feature, which isn't satisfied by standard escrow?
<__monty__>
Yes, this breaks down when the dev and the maintainer are the same person. That doesn't make it useless in general though.
<joepie91>
it means that this solves exactly no new problems since there is a trivial attack that defeats it
<gchristensen>
and I'd rather just give them money
<__monty__>
gchristensen: You don't have to trust a single person to do the escrow because of the decentralization.
<gchristensen>
yeah but if an escrow company is stealing they'll be gone soon enough
<__monty__>
And if a maintainer cheats people won't support their project.
<gchristensen>
great, so no need for the blockchain
<__monty__>
The maintainer gets way less cashflow because there are many so he doesn't gain as much by cheating.
lassulus has quit [*.net *.split]
dmc has quit [*.net *.split]
jcrben has quit [*.net *.split]
noefk has quit [Ping timeout: 240 seconds]
noefk has joined #nixos-chat
jcrben has joined #nixos-chat
<joepie91>
srk: hm, unfortunately limited to opencollective so far
<joepie91>
otherwise an interesting idea
<andi->
The whole "fixed with blockchain" issue isn't that great.. So hyped... I mean it only distributes more state but doesn't solve any guarantees. You would need a crypo based approach where you get really compromising material of the escrow PEOPLE (not the company..) when they cheat on you.. Not sure that's really feasible tho..
<__monty__>
I think most people miss the point. All the blockchain gives you is decentralized consensus. That's HUGE. However, it's not a panacea.
<joepie91>
is supposed to give you*
<joepie91>
in practice PoW turns out to basically destroy the planet if left unchecked, and none of the proposed alternatives are meaningfully centralization-resistant :)
<joepie91>
don't think we're quite to a point yet where we have reliable non-planet-destroying decentralized consensus
<gchristensen>
if the earth didn't want to be destroyed it should have been built on the blockchain with a smart contract preventing it
<joepie91>
would be nice if we did, but alas
<__monty__>
I'm not convinced the problem isn't blown out of proportion. You always hear about how much CO₂ "the blockchain" causes. But you never hear how much CO₂ conventional currency creates.
<infinisil>
Or cows!
<__monty__>
Long term it'll settle down either way, and at least it's putting a spotlight on china's horrible policies concerning energy.
<infinisil>
(okay no I'm definitely hearing about that, otherwise I wouldn't know that they produce so much of it)
<joepie91>
__monty__: thing is... you *do* hear about that, and people have run the numbers, and it's considerably less than what the Bitcoin blockchain uses; the problem is that the original economic rationale behind PoW chains - namely, people will continue to optimize the mining process as far as possible - turned out not to be quite true, instead working out to "people will spend almost as much on energy as they can earn from the mining process"
<joepie91>
in other words, there is no reason to believe it'll "settle down"
<joepie91>
it would only get worse as the perceived value of $whateverCurrency increases
<__monty__>
Not if CO₂ emission is taxed proportionately. It'll stimulate decentralization.
<__monty__>
Also, doesn't cardano already have working PoS?
<joepie91>
every PoS mechanism I've seen is prone to rapid centralization by adversaries
<__monty__>
Doesn't mean it's an impossible problem to solve.
<joepie91>
as for taxing CO2 emissions; do I really need to explain how it's ludicrous to expect (local!) taxes to fix an issue introduced by a poorly-designed consensus system, producing a lot of collateral economic issues in the process?
<joepie91>
__monty__: it also doesn't mean it's a *possible* problem to solve.
<joepie91>
and so far, every attempt to discuss this has played out as "but don't you see the POSSIBILITIES, they are endless!"
<joepie91>
rather than any rational evidence that a solution may be in sight or even possible
<joepie91>
so I don't take any of these mechanisms seriously until somebody addresses all the concerns head-on
<__monty__>
It's not up to local governance to solve the consensus problem. Things like kyoto agreements already exist. Countries are already incentivized to reduce CO₂ emissions. It's just not strict enough yet.
<joepie91>
compounded by the fact that those yelling the loudest about the supposed benefits of blockchains, seem to understand the least about their mechanics and economics
<joepie91>
as a general rule of thumb
<__monty__>
That's true of any technology though.
<joepie91>
hardly.
<joepie91>
the degree of hype surrounding blockchains and their derivatives is absolutely an outlier.
<joepie91>
to the point that basically every major well-reputed 'blockchain' project either isn't a blockchain at all, or has REALLY obvious issues
<joepie91>
Bitcoin had an excuse, being experimental and the first of its kind
<joepie91>
the rest does not
<__monty__>
I think it's interesting and I keep an eye on it. I'm not gonna denounce it until there's absolute proof it'll work. My experience with program verification has seriously relaxed my "I need to see proof." personality.
<__monty__>
Sure most of the projects are just trying to capitalize on greed.
<joepie91>
I mean, I've actively reviewed many blockchain-y projects
<__monty__>
Ethereum's an interesting evolution of bitcoin though.
<joepie91>
it usually doesn't take more than an hour to find the first glaring issue, and that's including the time it takes to read through the intentionally-obtuse mathspeak
<__monty__>
And some of the anonymous coins are interesting too.
<joepie91>
__monty__: ethereum has severe flaws in its whitepaper.
<joepie91>
it also changes very little compared to Bitcoin in terms of mechanics.
<__monty__>
Sure it does. It still has interesting ideas.
<joepie91>
it's essentially "Bitcoin, but with more possibilities for scripting"
<__monty__>
I know, I've read both the bitcoin and ethereum papers.
<gchristensen>
#nixos-discusses-blockchain
<gchristensen>
(gentle suggestion)
<andi->
nixcoin!
<joepie91>
these things just aren't interesting to me because of the glaring issues and the absolute disregard of the past twenty years of decentralization research and many more years of economics research
<joepie91>
and I don't find them to meaningfully contribute towards technology as a whole; in fact, I've multiple times had it actively interfere with my research
<__monty__>
I can't shut up. I have a compulsion to discuss things. So you'll have to gather supporters of the suggestion ; )
<__monty__>
What's your research?
<joepie91>
__monty__: the specific research topics that it's interfered with are 1) decentralized systems (in particularly, non-economic ones) by having basically killed off research into non-blockchain decentralization mechanisms, and 2) cryptography in general, in no small part due to the hijacking of the term "crypto"
<__monty__>
I see bitcoin as the first mostly complete solution to the byzantine generals problem. That's a huge accomplishment. Sure ethereum comes nowhere near that big a step. It's still interesting.
<joepie91>
like, beyond 2011 there is basically _no_ research in decentralized networks that doesn't add a blockchain somewhere
<joepie91>
sure, and it's the same reason Bitcoin interested me
<joepie91>
it's the absolute torrent of shit that came after it that I have an issue with :)
<joepie91>
and being completely realistic, Bitcoin was always an experimental project *at best*, and even explicitly stated as such
<joepie91>
but apparently experimental projects and money don't mix
<joepie91>
and it spun off into a tornado of money-hungry pseudo-inventors
<joepie91>
drowning legitimate research...
<joepie91>
(both in discoverability and funding)
<joepie91>
I think Bitcoin *was* a useful step in decentralization research; but that is where it should have stopped, and where people should have gone back to the drawing board to fix the issues that became apparent
<__monty__>
Like many other things, the bubble'll have to pop before the tech'll be used properly probably : /
<joepie91>
(yielding to other avenues of research, specifically)
<gchristensen>
> Miners spend $17 million a day for a shot at $4.4 million of bitcoin
<{^_^}>
error: syntax error, unexpected $undefined, expecting ')', at (string):171:14
<gchristensen>
if we spent 1mo of that money we'd have fixed Flint Michigan by now
<__monty__>
gchristensen: That's not exactly how it works though. Only if they sell the bitcoins they mine rn.
<gchristensen>
oh that article is 5yrs ago
<gchristensen>
(I'm referring not to the 4.4mil return, but isntead the 17mil spend)
<__monty__>
Same probably applies today since bitcoin's not worth much rn.
<__monty__>
Relatively.
<__monty__>
Problem is people will only spend money if they get something in return.
<__monty__>
Tbh Flint Mich. is a far from my bed problem. I don't even know what the problem *is*, let alone care about fixing it.
<gchristensen>
(that is a bad look, and signals it is time for me te loave)
<gchristensen>
(that is a bad look, and signals it is time for me to leave)
<__monty__>
Not knowing about Flint Mich.?
<srk>
I hope cardano succeeds.. built on good stack :D
<joepie91>
__monty__: they've apparently been without clean water for 3.5 years now.
<__monty__>
joepie91: All that makes me think is Africa still has it way worse.
<__monty__>
I'm not american so I don't have the America First mentality.
<gchristensen>
wow
<andi->
There is enough stuff that should be fixed on this planet.. no point in arguing what the first should be.. just DO SOMETHING
<gchristensen>
+1
<__monty__>
Not arguing, just sharing why I don't care much at all about Flint and why I don't think that's terrible of me.
<joepie91>
"As a result of this work, today we are releasing the Beta of a new and improved version of Steam Play to all Linux users! It includes a modified distribution of Wine, called Proton, to provide compatibility with Windows game titles."
<andi->
games \o/
<joepie91>
"Proton, the tool that Steam Play uses to provide Windows compatibility, contains a custom version of Wine as well as additional libraries developed alongside it. It's fully open-source and available right now on GitHub[github.com]!"
<andi->
Since I stopped playing games my pain in the right hand disappeared. I blame the mouse :)
<samueldr>
andi-: use a controller *ducks*
<srk>
I remember I've had to run system shock 2 with system wine instead of the bundled one recently to get sound
<gchristensen>
wow!
<joepie91>
DXVK seems to be the critical component
<samueldr>
the good news is that it's presented as linux-playtime for the developers
<srk>
looks like I can soon reclaim my windows gaming disk
* andi-
installed Star wars BF2...
<andi->
it launches, it looks good it just doesn't accept any input :/
<srk>
hehe, last week I've compiled out of tree ppm decoder module and created patchelf expression for fpvfreerider just to find it crashing somewhere deep in mono. but it launches, creates window, inits..
<srk>
need to send stacktrace to author
<andi->
ahh that reminds me I still have that sndio expression in my overlay that is required by a few games that provide native linux support..
<srk>
not sure if it's a unity feature or something custom but it also tries to run gdb on itself to get better stacktrace but that fails horribly on nixos
<srk>
even when gdb is in path and some debuginfo present
<andi->
argh such time waste.. back to doing something productive :D
<andi->
I simply don't know enough haskell to fix the compilation of the new elm release.. On unstable it failes because of `base<4.11` on 18.03 it does compile time code and tries to create random directories but can (almost) be built using cabal2nix :/
<jD91mZM2>
If anybody is missing my constant status updates, I'm almost rewriting rnix... again...
<andi->
jD91mZM2: oh, what makes you do that? New findings?
<jD91mZM2>
andi-: I've had a lot of contact with matklad who works for IntelliJ. He's been giving me tips for how to represent the AST structure in the most optimal way for editors
<andi->
nice
<jD91mZM2>
Current rnix is strong typed. Everything has a type an values. Iterating all the children of this node is hard, because it's not clear what is a "child"
<jD91mZM2>
He suggests making ASTs have a generalized children field
<jD91mZM2>
This way I can iterate children easily, and it's going to be a lot nicer to use to create tools
lassulus has joined #nixos-chat
<jD91mZM2>
It'll actually be more efficient also! All children can share vector with the main arena-based tree, meaning I only ever allocate one single (but big) vector
<andi->
wouldn't you be able to fix that with a bit of trait and generics?
<jD91mZM2>
Not with the current enum structure
<jD91mZM2>
How familiar are you with rust enums?
<andi->
I had a similar problem lately where I had to duplicate all of the enum fields and basicallya dd the same code toe each of them :/
<andi->
I was thinking of writing a macro for it but didn't continue that project yet
lassulus has quit [Client Quit]
<jD91mZM2>
Basically my problem, but mine here was even worse because everything had different fields
<andi->
srk: I just started learning elm. wait a few days before you throw the next best thing at me :P
<andi->
jD91mZM2: there MUST be some rust-ish way to solve it without duplicating lots of code :/
lassulus has joined #nixos-chat
<jD91mZM2>
andi-: That's what matklad solved!
<jD91mZM2>
Basically, each AST node now has a vector (not literally, it's a bit more efficient) of children
<srk>
hehe, wanted to use elm at first (really like its time traveling debugger) but miso is pretty much the same concept with familiar language :)
<srk>
ghcjs is mad
<jD91mZM2>
You no longer say `entry.name`, you loop through the set's children and find the first identifier.
lassulus has quit [Client Quit]
<andi->
srk: I haven't completed a haskell hello-world in my life yet.. So I stick to the "easier" (from the looks) language for now. I am really trying to learn more functional programming.
<jD91mZM2>
This is a little weird, but it solves the problem of duplicating code, and it's a lot more efficient too because since EVERYTHING has a vector of children I can use my existing arena
<LnL>
thought about that too, but I think this might be a bit better since the has is determined but the unpacked content
<LnL>
fetcharchive doesn't really indicate that in any way
dmc has joined #nixos-chat
<gchristensen>
imo I'd leave it as-is
<LnL>
but I'm fine with another name if it seems better
<gchristensen>
"Why a fetcher just for a zip?" is a pretty easy question to ask / answer with experiments and validation
<LnL>
leave as-is as in fetchzip or fetchunpack?
<gchristensen>
fetchzip
<infinisil>
Why don't we make a general `fetch` that works for pretty much everything?
<gchristensen>
we do, a fixed output derivation
<LnL>
fetchurl doesn't do anything with the archive, you only want to unpack before hashing in certain cases
<LnL>
so what's your thought there? do you think the name is fine or that it's not worth causing potentially even more confusion by introducing another name
<gchristensen>
I think a combination of both
<gchristensen>
the name isn't ideal
<LnL>
I agree with the second part and I'm not sure with the best approach is there, but I think the name is pretty bad
<joepie91>
narrator: the system was not stateless
<infinisil>
State, state everywhere!
<LnL>
kill the state!
<joepie91>
LnL: you just ended up on a list
<joepie91>
either a list of functional programmers, or a government watchlist
<LnL>
lol
<gchristensen>
why not both
<joepie91>
I think I left a data structure joke on the table there
disasm has quit [Ping timeout: 260 seconds]
<joepie91>
something something list...
<LnL>
you forgot unit at the end
<joepie91>
gchristensen: hmm, a government watchlist of functional programmers...
<joepie91>
I now have even more questions :D
<gchristensen>
radical anti-statists
<srk>
this
<gchristensen>
and their fringe splinter group which goes by the abbreviation "IO"
<joepie91>
hah
<gchristensen>
can I get help writing a sentence?
<srk>
I'm trying to create a haskell version of nixops but I'm not sure about the state and effects atm
<joepie91>
gchristensen: are you looking for a way to rephrase it?
<samueldr>
gchristensen: ?
<gchristensen>
yes
<joepie91>
gchristensen: I do have some ideas, but what do *you* feel is insufficient about the current phrasing?
<samueldr>
the american asks the ESLs (assuming joepie91 isn't a native english speaker)
<joepie91>
samueldr: you're assuming correctly :)
<samueldr>
:)
<gchristensen>
it is pretty negative I think
<joepie91>
samueldr: most people assume otherwise though :P
<gchristensen>
instead I'd like to be positive and say it is so cool you can install anything since 1.11.16
<joepie91>
right
* joepie91
processes
<infinisil>
srk: You can even put all IO actions you need into a typeclass, then only use that typeclass and make IO an instance of it to run it
<infinisil>
srk: And you could then use a different implementation of that typeclass for pure testing
<samueldr>
Starting with version 1.11.16, all nix versions can be used
<srk>
infinisil: yeah, trying to figure out exactly this thing
<samueldr>
or all nix releases?
<infinisil>
srk: I have painfully figured it somewhat out some time ago ugh
<infinisil>
Monads, monads everywhere]
<infinisil>
!
<joepie91>
gchristensen: "This option supports all versions starting at 1.11.16. Older versions are not currently available."?
<joepie91>
possibly s/currently/yet/
<joepie91>
depending on how much you intend to add other versions later
<gchristensen>
that is nice
<srk>
infinisil: I saw that in hnix and hnix-store, iirc iohk has soem good docs
<samueldr>
this focuses on a lower bound and is open
<gchristensen>
very little :) it was only a "nicety" that we added support back to 1.11.6
<joepie91>
gchristensen: also, are you interested in some of the reasons why the old phrasing doesn't have great readability? or is just having the new phrasing sufficient?
<joepie91>
then I'd stick with "currently"
<gchristensen>
16*
<joepie91>
as "yet" strongly implies a near-future change
<gchristensen>
not especially :)
<joepie91>
alright :)
<gchristensen>
how about "Note: This option supports all Nix releases, starting with 1.11.16."?
<joepie91>
I'd probably add 'version' before the number, otherwise that looks fine to me
<joepie91>
(the addition of 'version' primarily being to help people who are scanning specifically for version info)