Hacker News Reader

Comment by jrockway | Hacker News Reader

I feel like the implication is that people prevent you from customizing stuff for ego reasons, but isn't it a lot of code and testing to implement the personalization UI?

But thinking about it more, I suppose the answer is to let the users write their own clients. Netflix should have a JSON API that's a list of all the programs, and each record should contain a link to video file. They don't do that because of DRM and they specifically want to prevent you from writing a program that saves the downloaded bytes to disk. So I guess there is no way to achieve their business goals (make sure you pay $20/month forever to watch stuff, and make sure the "rights holders" get to sell you the same thing multiple times) without doing more work, which they simply chose not to do because they don't have to.

Going one level deeper, I guess that video streaming is a weird market. You could have the worst software and the best content and probably win. So you can't compete with Netflix by making a video streaming service that has more UI features. I guess what this means is that copyright is deterring innovation. I think I've heard that one before. Sigh!

View on HN

> but isn't it a lot of code and testing to implement the personalization UI?

But thinking about it more, I suppose the answer is to let the users write their own clients.

That's certainly one reason (on the operational layer), but I very much doubt it's the main reason. Otherwise, companies wouldn't go out of their way to prevent you from writing your own client.

My suspicion is, there is an additional layer between data gathering and customisation that is extremely interesting for companies: the target metric.

If you're allowed to customize things yourself, you supply your own target metric to judge if the current settings work well for you. It's mostly implicit, but I'd guess in most cases, it would be about ergonomics, how easy you can archieve things that are important to you, how much it matches your personal visual style, etc.

Telemetry and "forced personalisation" allow the company to supply their own target metric instead. Yes, ergonomics and personal preferences play a role, but the company can easily mix in other things, such as retention, which are exclusively important for them but not for you.

There was an article on HN the other day that suggested that (and I'm paraphrasing) "only failed products have settings".

We hear a lot about simplicity and how it's easier to study users and optimize for desired outcomes: conversions, upsells, continued use, whatever. Having a simple product nudges the user in the direction you want them to go and makes your product intuitive.

I think there's another reason for simplicity: incompetence. Your team doesn't care enough about the product that they're actually using it. There are no in house power users. They're not making it for themselves, so they simply do not care. They're there to finish the Jira tickets and shovel out a Fischer Price product that meets the requirements doc.

You can tell which software was created by power users. Vim. Linux. Blender.

> Vim. Linux. Blender.

All of which are famous for having user interfaces/experiences that aren't very intuitive for the newbie. Which is okay, but perhaps different segments of the population have different needs and therefore are best served by different software.

> Linux

To be fair a couple of mainstream distros are far easier to get than Windows or Mac, only people have a lot of previous Windows experience by the time they encounter Linux.

These days iPads of course are at a different level, but that wouldn't be a fair comparison.

Source: I had a 6 years of power user experience from Windows when I first started using Linux.

>To be fair a couple of mainstream distros are far easier to get than Windows or Mac

Until you try to plug in a WiFi Dongle, printer, SD Card reader or pretty much any other piece of hardware that isn't a keyboard or mouse.

Those all mostly work out of the box these days. Even devices that don't get driver support on newer Windows anymore.

Depends on your definition of "work". I've had issues with every device I listed on Kubuntu 20.04.

I, on the other hand, had an issue with every type of device you have listed using Windows, but none on Arch Linux.

And I can add to the list devices like: an SATA to USB adapter, gamepad (not only DualShock, but also X1 Gamepad [sic!]), external CD-ROM, Wacom tablet

With each of them there was an issue on Windows, but on Linux it's Plug-n-Play; heck, even my gaming mouse needed drivers on Windows for all buttons to work, while on Linux noting additional was required

A while ago I bought a used Nikon Coolscan negative scanner. The last edition of those came out in around 2004 and the last driver only supports Vista.

It came with a PCI Firewire card. My installation process on Arch was installing the card, rebooting plugging everything in and installing SANE. That's it, I could now scan my negatives.

It's very hardware-specific so it's impossible to generalize, but usually stuff will either work out of the box or not work at all on free systems (Linux, BSD). The exception is if your distro refuses to ship with proprietary software (good!) in which case you need to install extra drivers (eg. firmware-linux) for stuff to work.

Something to consider is how old the hardware is. Typically, drivers will land in Linux kernel 1-2 years after hardware is released, so you need to use slightly older hardware and/or a distro which keeps up to date on a rolling-release basis. For example, good luck finding 802.11ax hardware supported by free systems right now: 802.11ac on the other hand (spec from 2013) works perfectly for 90% of chipsets.

To be clear if you're not familiar with driver development: this is 100% the fault of crappy hardware manufacturers who are doing everything except their job. Their job is to produce hardware and publish datasheets for them: problem is due to corporate/copyright culture and abuse of dominant position from Microsoft since 90s (see IBM/Microsoft scandals and the threats to dismantle Microsoft), they stopped to publish datasheets and instead publish a crappy windows driver that's full of bugs and will stop working with new releases.

We should not be supporting these vampires with our money but hey the problem is the hardware vendors who do distribute their datasheets can be counted on our fingers and their hardware doesn't have the raw power of your latest NVINTELDIARM.

Excellent usenix talk on this topic: https://www.usenix.org/conference/osdi21/presentation/fri-ke...

Previous HN discussion on this topic, about Lenovo more precisely: https://news.ycombinator.com/item?id=28977175

Since around 2008 most of the stuff I use have been better supported on mainstream Linux than on Windows.

Literal plug and play.

>All of which are famous for having user interfaces/experiences that aren't very intuitive for the newbie.

Does any other vim alternative have less features and settings? Notepad is not an alternative. I am not a vim fan but my text editor and my IDE are power user tools and have settings.

Linux, probably the OC is referring to the kernel, otherwise GNOME is famous for it idiotic simplicity where you must modify your work to fit the software and not adapt your server to fit your work.

Blender , what does Maya (or other alternatives) has less settings and menus?

I think you missed the point, pro users are not all identical, working all n the exact same way on the exact same things, your application that targets professional needs to adapt or reject from the start a big part of the user base.

> Does any other vim alternative have less features and settings?

As a long-time vim user who doesn't use the power features (except search/replace patterns) i've found micro to be a very decent alternative to vim/nano to teach to people new to CLI land. Select text with shift, save with ^S, quit with ^Q, find with ^F... it's just like windows notepad, but with colors/indentation out of the box.

There were also some conversations a few years back in emacs mailing list about the choice of defaults. Some people argued emacs defaults should try to accomodate people who've never used emacs before but are familiar with other key bindings and UI paradigms, because power users will always find how to configure their favorite workflows but newbies will leave if they can't accomplish the most basic tasks without reading a 50 page manual.

You don't need a default war, you hopefully have intelligent enough users so you can show them a popup where they can select a profile that is more comfortable for them.

I hated vim so I can't say too much about it, just that I hate when it is set as default on some servers (or some colleague thought is cool to make it default) but I would not attempt to attack the vim project or vim users for their choices, I understand that it is perfect for them and we need hyper-complex-text=editors too, it would be stupid if Notepad users will start writing blogs that say Notepad++. vim and emacs are complex because the developers or users are stupid and are using Notepad wrong.

The reason I stopped hating vim is because they added instructions on how to quit it, so I can quit it and rerun my commands with EDITOR=nano in front when vim pops up unexpectedly.

Are we really at the point where having a button to disable personalization makes you a "power user"? I'm by no means a twitter power user, but I do want to just make my timeline chronological and remove all the "tweets you might be interested in" garbage.

This! I did not explicitly ask for my entire feed to contain "threads" of startup advice from random people.

One big problems with "settings" is it's easy for people to change them, then get confused.

You used to be able to move the taskbar around the screen, and make it taller / shorter. I'm sure some people loved this. On the other hand, I would often find people whose computers were "broken", as their taskbar was invisible, or took half the screen.

Even experts can fall for this -- I've had a few open source projects where someone complains "I did X from the documentation, but it didn't work". After a while we discover they, months ago, switched off X (or something related to X).

I'm not saying there shouldn't be settings, but they aren't a clear win. Even ignoring the fact they all need testing, they provide pain points for many users.

Yeah, I think the real problem here is that nobody wants to support anything anymore. The only way to continue shipping products is to basically ship generic throw-away happy meal toys that are incapable of being misused.

They could do that but you deeplink back into the Netflix app to actually watch to solve the DRM issue. Which I think shows that it's about more than just DRM. They want to control the experience.

Of course they do.

If they ever have any hope of improving margins, they need to argue that “our curation drives content discovery” to get better deals out of content providers.

they've abandoned the content acquisition strategy. that went out the boat a long time ago. netflix is a content creation company now, competing with HBO, NBC, etc.

That doesn't change the value they provide to the people actually making the shows.

As a producer/show runner, who would you rather make a series for, Netflix or a streaming service you've never heard of? Which would you accept less money upfront from?

Suppose you choose Netflix to fund your philosophical sci-fi show? Netflix tell you they will put it at the top of the main menu for everyone who's ever watched the new Star Trek as long as you agree to X. Do you agree to X?

Thank you. It's always a struggle to communicate to people in 'labor' industries that creative content is ruled by 'talent,' so even if you employ the 'talent,' they still have the upper-hand as the owners of their IP and abilities.

Importantly, to get the data to learn from it to improve faster than their competitors can.

We realized our competitors were copying us at a company long ago. We geofenced a/b tests so they wouldn’t see what we were testing and intentionally made the web app worse for that area (mostly as a “fun” 20% type project). Interestingly, the “worse” version starting doing really well even though we were giving them “control” for all older features. We ended up rolling back several months of changes to the entire product and we saw a huge burst in revenue.

It wouldn’t be until years later until I understood the Simpson paradox and that a/b testing improves the mean of whatever metric you’re trying to improve, usually by decreasing some important segment and improving some unimportant segment.

That's why you should measure many things - especially all your important things - in every a/b test [=

With the Simpson paradox, even if your metrics go up, across all your segments, it’s actually going down. Measuring only takes you so far.

Sure - this is why Google Search constantly makes less money. Measuring only takes it so far.

I would love to read a blog post with more details

That description of Simpson's Paradox doesn't really agree with my understanding. Simpson's paradox goes with how the same data is cut into subgroups and can invert the numbers.

That is, it's less about "metrics up, other unmeasured metrics down" and more "we're up in all subgroups, so we never noticed that our numbers across when we combine the subgroups are down."

There's no practical difference between the two.

DRM doesn't effectively keep you from finding everything anyway, and much DRM is flimsy, easy-to-bypass pro forma DRM; because DRM -is- about controlling the experience, first and foremost.

Its sheer presence activates the DMCA felony protection of circumventing copyright. This is why Cory Doctorow always calls copyright evasion "felony contempt of business model". It's there so the company can control the experience and squash any attempt to bypass that control; in that sense, the DRM on the UX is more important than the DRM on the content itself.

Obligatory Cory Doctorow "Unauthorized Bread" link: https://arstechnica.com/gaming/2020/01/unauthorized-bread-a-...

People who create and build things tend to have strong emotional needs to control that thing. This includes for example, the ui designers at netflix, and the people who run the company. You see this in all kinds of domains from filmmakers who don’t want their movie released on certain formats even though fans want it, to game designers nerfing beloved tricks because content gets skipped, and making cutscenes unskippable, apple wanting control of what apps you put on your phone and how you repair it and so on.

Often there is a profit motive as well but I think the underlying prime cause of this stuff is just control

Netflix has some vision for how their service should be used, and they don’t want you fucking with it, and various weird contractual obligations and profit optimizations are gonna get you less and less choice.

The people who want to customize are a small minority so there isn’t any competitive advantage to offer it, and niche players can’t exist in domains like netflix, or iphone.

niche players can’t exist

This is where the world has gone astray. We would be much better off with a thousand three comma companies taking the place of a single four comma company.

Also, niches used to be considered socially acceptable. What has taken all the air out of the room?

Sure but in domains like running a streaming tv service, or building a phone, catering to a small subset of users means you are a smaller company, and then it is very hard to build a decent phone, or create something like netflix.

But why is that? Phones are more expensive at retail and more profitable than ever. Why should it be so difficult to make one even with a billion dollars? And it wasn't hard to run a video rental store 25 years ago. What changed?

Phone profit margins have been declining even for Apple (about ~60%) and are already fairly thin for the rest of the market (~13%). The problem is that getting the kind of customization we're talking about here probably means rolling your own phone OS which is a non-starter.

[0] https://www.forbes.com/sites/johnkoetsier/2019/12/22/global-...

The only must-have feature from a given new phone for me is the upgraded camera. In most other regards, reusing stuff from previous generations where profitability peaked should be fine, giving a dozen niche manufacturers enough clout to build customizations off a common base.

> They don't do that because of DRM and they specifically want to prevent you from writing a program that saves the downloaded bytes to disk.

This seems like a 20th century philosophy.

The point of a subscription service is to watch whatever you want. This one came out last week. You didn't know that one existed until today. You can't have downloaded it before it existed or before you knew you wanted to watch it. So there's no point in even bothering to do this.

And if people actually want to pirate your stuff, the DRM empirically isn't stopping them anyway. It's just a pointless farce.

> You could have the worst software and the best content and probably win.

There's winning and then there's winning.

You're not going to get any subscribers with no content, true. But Netflix has the content. Now suppose they had a better interface. Better experience, people spend more time watching Netflix, Netflix is more valuable and can charge more or gets more subscribers. It's not useless just because it's not the dominant factor.

Or maybe they found a balance between ppl watching just enough to keep their subscriptions and not get bored of Netflix.

Just thought about this and it made me a bit scared tbh. :D

> And if people actually want to pirate your stuff, the DRM empirically isn't stopping them anyway. It's just a pointless farce.

Indeed, if the past 30 years of digital piracy have taught us anything it is this. It should be obvious: it is impossible to both allow consumption of something while simultaneously preventing its duplication. All DRM does is make life worse for your customers.

> You didn't know that one existed until today. You can't have downloaded it before it existed or before you knew you wanted to watch it.

Why not? If Netflix made their catalog DRM-free and available via an API, then you could download all of Netflix, and then unsubscribe. Sure, you’re missing out on the new stuff, and it could take a lot of disk space to store all of Netflix, but you still have 6k titles [0] to watch for $20 (assuming you’ve got an ultra fast connection and manage to download it in a month, of course). And if you somehow run out of content (over 36k hours of it, according to a slightly older source [1]), another $20 will let you expand your library.

[0] https://www.whats-on-netflix.com/news/netflix-library-by-the... [1] https://www.whats-on-netflix.com/news/how-long-would-it-take...

I've wanted that model for music streaming for a long while now. Let me pay for a streaming API key and plug it into third party software that doesn't have a flavor of the day interface and set of features.

Apple Music does this.

Is the protocol documented anywhere?

https://developer.apple.com/documentation/applemusicapi

How does that let me stream the actual song? It provides access to metadata only, as far as I can tell.

https://developer.apple.com/documentation/musickitjs

I think you can use it in conjunction with this JS library: https://developer.apple.com/documentation/musickitjs/musicki...

Musi.sh my friend.

> You could have the worst software and the best content and probably win

I definitely get where you're coming from. As a counter example, my parents aren't great with tech and have written off platforms with content they want because they find them annoying. For example, they find Amazon Video way too confusing. What's free? Why are there channels that aren't free that they already have somewhere else?

> I suppose the answer is to let the users write their own clients

Netflix definitely doesn't want this. They want people to turn on their TV and open Netflix. If people have a third-party client that can show Netflix content alongside content from HBO, Amazon, Apple, and Disney, people might start watching their competitors more.

Not only that, but Netflix wants to be able to push your viewing in certain ways. If everyone starts watching a show at the same time, it can create buzz. Netflix's big promotion of Squid Game got to that critical mass for buzz to happen.

If they're able to determine what comes up where in the UI, they can also prioritize shows that are cheap for them. Getting someone interested in an owned show is better than getting someone interested in a show that they're renting from a third party.

That said, there are third party UIs like JustWatch that allow you to integrate with multiple services. We also saw Movie of the Night (https://news.ycombinator.com/item?id=29862588) featured here a little while back offering an API and it does offer links to shows in Netflix.

I think the issue is that it takes work to build a third-party UI and it takes money to actually collect the data (or pay someone else for the data) while the streaming companies aren't totally happy about giving you that data.

Really, everyone is looking for an edge. Google wants their Android TV to serve you ads for shows. They want Hulu to pay them money for them to push Hulu shows in their UI even if you don't have Hulu. Roku wants to sell advertising space on their Home Screen. Every content provider would love to become your primary destination while you ignore the other apps on your system and forget that they have content too.

> but isn't it a lot of code and testing to implement the personalization UI?

Most of the stuff cited as example come from additional features changing an existing behavior. Allowing to disable it would be overhead, but really not much I think.

Another example: Youtube started to automatically translate “foreign” video titles. Having a “never translate anything ever” toggle in settings would smallish, there is no other UI impact, and they must already have that kind of setting internally for testing or legacy purpose.

>You could have the worst software and the best content and probably win.

You should see the apps from commercial broadcasters in Australia. Last time I tried using 10play/7plus (to be fair, this was 2 or so years ago), I swore off of using them ever again.

No progress tracking, batshit insane navigation, episodes playing out of order and there was a bug where it would semi-regularly not even load the right video.

Of course, good content can make up for some minor software annoyances, but if the experience is bad enough the content doesn't even matter any more.

And that's why video broadcasters should be mandated by law to follow standards for interop: whether it's HTTP+HLS over a specified API, or simply FTP/NFS. When you think of it, it's crazy how backwards we've gone now that you need a netflix client to play netflix videos and a youtube client to play youtube videos: it's like VLC/mplayer never existed!

To be fair, it's possible to use 3rd party youtube clients, but Youtube changes its pages/APIs just often enough (~ once/month) to break everyone's apps which is evil.

"I suppose the answer is to let the users write their own clients"

Have you tested that supposition against the percentage of users that can write software and compared it to the percentage of users that don't code for a living?

Well first yes it's been tested: that's how email, DNS, WWW, XMPP, Matrix (and others) work. Heck, that's how most of the Internet works!

But also, maybe the problem is we often develop apps as a fragile monolith which is definitely not intended to be customized in terms of UI. When the core or your application is designed library-first so that your executable is only an opinionated wrapper (client) around the lib, you make it easier for others to reach their usecase by implementing a different wrapper. I don't see a reason this wouldn't apply to composable/reusable GUI components: early Mozilla (XUL) is a famous example of that.

>I suppose the answer is to let the users write their own clients

What happens when Netflix wants to release a new feature or change an API? Do they now have to worry about getting everyone else to update their clients? What if some customization is put behind a paywall. What should happen if a third party client gives it away for free?

Reply on news.ycombinator.com