Fork me on GitHub
#off-topic
<
2018-04-01
>
qqq03:04:10

Is there a way for website operators to get paid micropayments via having readers solve mechanical turk tasks?

seancorfield04:04:47

Haven't we had this whole micropayments discussion?

👍 16
qqq04:04:43

Possible, but I can't find it, and I don't remember the resolution.

qqq04:04:52

What was the conclusion?

seancorfield05:04:00

@qqq This is an unsolved (and potentially unsolvable) problem. And probably not a problem any realistic company cares to solve.

ajs08:04:33

I read an article about a month or two ago about a not-tiny company who is doing exactly this — you can either subscribe to access their site (for a fee) or use it for free and they get to use your idle computing power to mine bitcoin. I think it was posted to HN. Commenters there seemed to think it was a potentially good idea.

seancorfield05:04:30

Again, you are obsessed with a "problem" that no one else seems to want to deal with.

dominicm06:04:33

Python stacktraces < Clojure stacktraces There's not even a message!

dominicm06:04:03

There usually is, but not when tests fail. Unless the test is checking a list. Woah, weird.

mpenet08:04:21

@seancorfield what part is unsolvable? Micropayements (without fees) is a (partially) solved pb already, at least outide of the US. Ex: swish. The rest seems doable/not too far fetched.

tbaldridge14:04:03

The discussion that we've had several times is that @qqq wants some sort of system like StackOverflow where people can pay small amounts of money for working results to programming tasks.

tbaldridge14:04:51

There' a rather large list of reasons why this won't work, aside from it being a disaster in the end because your app would be written by 1000 developers all who are more concerned about getting a pay-out than actually creating a good design

👍 12
tbaldridge14:04:08

And that's the core problem: programming is more often about creating a good design, the actual algorithms and code involved are either: a) inconsequential or b) so hard to solve that you need an expert, the type of person that's never going to hang out on SO to be paid peanuts solve small tasks.

👍 12
tbaldridge14:04:07

Not to mention that micropayments are too small for this sort of thing, since most programming tasks would take me about an hour to finish, by the time I setup an environment, wrote the code, tested it, and submitted the result. So what, are we going to pay $3/hr for programmers? Pay people a decent wage for this sort of contract work, and you're looking at $75-$100/hr and at that point PayPal works fine, so you don't need micropayments

mpenet18:04:49

Right, the "task" part was a bit vague, if we're talking about programming. micropayements do not make sense indeed.

qqq23:04:52

@U050SC7SV @U07TDTQNL @seancorfield: IIRC for the "pay by the function" idea, I was referring to $50-$100 / function micropayments: generally refer to < $1.00 / transaction the question here, in mentioning mturk was intended to be the question of: suppose you're a new org, you want people to pay per article, but it's inconvenitn to charged ppl $0.01 / article they read -- so could you get people to solve $0.01 worth of mturk HITs instead?

tbaldridge23:04:03

>> "but it's inconvenitn to charged ppl $0.01 / article they read -- so could you get people to solve $0.01 worth of mturk HITs instead?" Nothing about that makes any sense.

qqq23:04:17

Right now, there is not a easy way for an online newspaper/magazine to charge a reader $0.01 to read a particular article. One possibility is: the website gets $0.01 worth of mechanical turk tasks from Amazon the website presents the tasks to the reader the reader solves the tasks the website gives the reader the article mturk adds $0.01 to the website's account

tbaldridge00:04:59

So I wanted to read an article, and instead I have to solve some puzzle? Is probably do what I do with any news paywall and walk away.

qqq00:04:30

That's fine; the issue is not whether they can get 100% conversion, but whether they can make more money.

seancorfield02:04:39

I don't think they'd make any money that way. Monthly subscriptions works for most companies. Buying credits to use later (in microtransactions) seems to work for a lot of others.

seancorfield02:04:11

People want a simple trade: money for benefits.

mpenet08:04:14

If it's a good/bad idea is another issue

tbaldridge12:04:33

@dominicm most Python test suites run in quiet mode where they capture all output during the test. There's a way to unsilence them. But I've always found Python test suites to be one of the better parts of Python.

dominicm12:04:26

I think I ran into a particular edge case, but it was my returning experience today. There was no "4 does not equal 1", just a stack pointing at the assertion line. If I did assertEqual [1] [4], it quite happily told me that they were different (similar to how humane test output does for clojure)

dominicm12:04:33

Hypothesis is really great though, I say that without comparison to test.check. I found 2 bugs in the python nrepl library which have sporadically cropping up for months. Both are now fixed. I also suspect the bugs I found previously would have been found by this process. For all of 15loc to test, using it on an encoder/decoder pair seems very logical.

tbaldridge13:04:57

Hypothesis is a strange beast, test.check does much better, imo

tbaldridge13:04:01

What test suite are you using?

tbaldridge13:04:23

Because python is interpreted, I've gotten very clear test result information in the past, I'll have to look up what I was using, but I've seen examples like this: failed func(x, 1, 1) == 42 where x == 40 where func = <some_module.minus>

dominicm13:04:27

https://github.com/clojure-vim/nrepl-python-client/blob/master/test.py#L49 it's possible I was doing something like self.assertEqual([i], next(…)) where next would return only 2 or similar

pesterhazy12:04:58

Does anyone know a guide, or guides, to JavaScript and its tooling, similar to @magomimmo's wonderful modern-cljs (https://github.com/magomimmo/modern-cljs) ?

sundarj21:04:25

dunno about any tooling guides

sundarj21:04:43

i recommend looking into Rollup: https://rollupjs.org/guide/en

pesterhazy12:04:56

I'd like to get up to speed with the recent changes to the language and its ecosystem

sveri14:04:34

Funnily I wanted to look into JavaScript today too. I was looking for a template for frontend and backend with typescript, webpack and react support. I only had AV few minutes so far, but it seems like my Google fuu has left me.

pesterhazy11:04:48

I’m curious too, let me know if you find something

pesterhazy11:04:29

create-react-app is popular, but not sure if it includes TypeScript

pesterhazy11:04:42

Ah looks like @lee.justin.m already mentioned it

dominicm16:04:12

How does one normally flag an option as required in a --help output? tools.cli doesn't have default formatting for it. Maybe the answer is that it should be a positional argument instead.

dominicm17:04:05

I just discovered http://docopt.org/ which not only answers that question, but goes from a standard documentation layout to a command line parser. Very cool.

justinlee17:04:54

if you want to take a look under the covers you can run it and then eject

sveri17:04:21

@justinlee Thanks, I already saw that one and it seems like a good start. What I am missing there is how to build a web application that serves from a backend and offers a backend API with node.js. I guess this is another topic, but it would be nice to have all-in-one guide.

justinlee17:04:09

@sveri in my experience people tend to treat the topics separately given that the defacto standard is a json/rest api, which is pretty decoupled. the node backend can be shockingly simple with an express server

justinlee17:04:38

it also may be that the backend requirements are so varied that there’s really no universal guide possible

justinlee17:04:01

note that during development, the c-r-a development server has a proxy feature so that you can run a separate node api server. then during production, you just serve the compiled bundle from the api server as a static resource

sveri17:04:23

@lee.justin.m I agree it may be difficult, but I am spoiled by clojure again, because if the existing templates that provide exactly that. Even with creatinga fat jar that contains both backend and frontend code. Would be nice if that existed for JavaScript and I guess it is possible in the landscape too.

justinlee18:04:51

you kind of just don’t need that stuff with node/express. it’s dead simple. its the one piece of my stack that doesn’t need a template, so I actually understand it and it is very predictable

sveri18:04:56

So it is dead simple to pick a template engine a routing engine, a db engine, a caching engine, maybe some useful libraries, maybe a key value store? When I start out new I would love to have that picked already as it saves me the time to research every single piece. Guess I am looking for a JavaScript webframework.

justinlee18:04:07

you are right that there is no rails like experience aside from meteor

justinlee18:04:05

also if you’re looking for templating and routing, create-react-app may not be what you are interested in because that stuff is spa oriented