Y HN SearchHacker News Search

new|comments|top|jobs

loading thread…

6 ms·

> Webapps are user-hostile What? What if I wrote it for my own use? What if it's among the thousands of web apps meant to be installed on a server you control?

by maratd 10y ago

> Webapps are user-hostile What? What if I wrote it for my own use? What if it's among the thousands of web apps meant to be installed on a server you control? There's nothing inherent about the technology that requires third party control.

CaptSpify 10y ago
I'm not who you were replying to but... I think web-apps are generally user-hostile. You are correct that there is nothing inherent about it though. I think the same could be said for FOSS: There's nothing inherently user-unfriendly about it, it just has a tendency to go that direction.
mindslight 10y ago
"Webapp" generally implies something administered by a third party. The extra overhead of (dedicated server, network setup, redundancy, each user requiring connectivity) only pays off when it's amortized over many users who do not have to set things up. Sure, some of these release their code as "open source" and you can setup your own copy. But I've never heard of one where the majority of users actually do so [0]. There are some locally-hosted applications that are intended to make themselves available on :::80 that use a browser as a UI toolkit, similar to how we can have a native app that links a browser as a library. But such things generally aren't referred to as webapps. The network transit seems integral to this term. [0] Well there's probably some developer-focused tools where this could be the case because said developers already have to deal with "web stuff" and so the additional overhead isn't as high. But this doesn't generalize to all users.
- maratd 10y ago
  > But I've never heard of one where the majority of users actually do so https://github.com/Kickball/awesome-selfhosted https://github.com/Kickball/awesome-selfhosted
  - mindslight 10y ago
    The majority of that list has nothing to do with the web (eg sendmail). And many of the projects that do also fail the requirements - for instance most users of a blog platform are readers, who still aren't in control of said software. We're obviously not disagreeing about the utility of local software, just that of "web apps". I assert that the platform itself is inherently biased towards making non-Free software. It's not impossible to create something that qualifies as a web app and Free software, it's just that this puts it in one small corner of an ecosystem based on user-disempowerment. Obviously many web technologies could be useful to build freer systems (like say if DNS/HTTP were replaced with something distributed), but as it stands the term basically refers to the modern generation of proprietary software. I get that a lot of by-day web developers want to apply their skills to build Free software, and that is great. They just need to be aware that it's an uphill battle against one's own assumptions, and building the same type of centralized systems desired by business interests (but releasing the code) likely does not suffice.
    - maratd 10y ago
      > The majority of that list has nothing to do with the web There are 188 PHP web applications in that list. 60 NodeJS web applications. 86 Python web applications. 25 Ruby web applications. I can go on. You're just being silly by pretending that this segment doesn't exist. > I assert that the platform itself is inherently biased towards making non-Free software. As opposed to native applications? I recall those being distributed through "app stores" where you have zero access to even the binary, nevermind the source code, and demand payment in return for access to either the application or features within that application. Does that mean native applications have a bias toward "making non-Free software"? No. And the answer is no for web as well. > They just need to be aware that it's an uphill battle against one's own assumptions, and building the same type of centralized systems desired by business interests (but releasing the code) likely does not suffice. I have no idea what this means, but based on your previous statements you're assuming a multi-user registration model with a centralized database you find on every commercial site which is a pre-requisite to monetization. You don't need to have that for a web app. At all. You can build a calculator as a web app. Or any desktop or command line application as a web app. With no registration or database required. Just type in a web address and start using the app. There is nothing inherent about web development that requires a user-based system. There is no uphill challenge to building free software using web. Now, you obviously won't have as many users of your software because getting things running is much more difficult ... but that's changing with platforms like Electron. http://electron.atom.io/ http://electron.atom.io/
- paulryanrogers 10y ago
  OwnCloud/NextCloud

mindslight 10y ago
As I keep saying, yes it is possible to create something similar to a webapp that is Free software. [0] But, let's take a look at the actual technologies behind the "web": 1. A naming system/transport protocol 2. A document markup language 3. A virtual machine for sandboxed code. (2) is user-friendly (PoLP). But creating a Free app implies ignoring the defining "web" features of (1) and (3)! If a single user is the only one accessing a given name, then that naming system has basically been made irrelevant. (If multiple users are accessing a given name, then per HTTP only one is in control of that name, and thus we deviate from free software). If a user controls updates to the code and (in theory) audits it, then the primary purpose of the sandbox (executing random third-party specified code) is gone. The battle-hardened sandbox is nice for security, but the functionality is otherwise indistinguishable from any other language. Why should such an app be called a "webapp" if it does not utilize most web features? > based on your previous statements you're assuming a multi-user registration model with a centralized database you find on every commercial site which is a pre-requisite to monetization I am not assuming this - I am noticing that a large majority of released projects are. Either due to lack of developer thought about repudiating "web" assumptions, or deliberate corporate "free-washing". Platform defaults end up defining the culture of a platform. A good question to ask yourself - if you were designing a Free calculator app and wanted to show a friend, would you send them the files you're editing and they'd set it up on their own httpd to try it out, or would you end up sending them a URL to your server and thereby undermine their freedom? [0] Although as I've said, many of those PHP etc apps are not actually Free - for instance the blog platforms.
- maratd 10y ago
  > If a single user is the only one accessing a given name, then that naming system has basically been made irrelevant. No. It's not irrelevant at all. It's how you "execute" that specific software. It serves the same purpose as a filename for a binary on a filesystem. > If a user controls updates to the code and (in theory) audits it, then the primary purpose of the sandbox (executing random third-party specified code) is gone. No. An audit is never perfect and that even assumes you perform one. While I frequently modify the software I utilize, I almost never audit the entire codebase, so a sandbox is quite useful. > if you were designing a Free calculator app and wanted to show a friend, would you send them the files you're editing and they'd set it up on their own httpd to try it out, or would you end up sending them a URL to your server and thereby undermine their freedom? I did just that. I sent them a URL. Here's the rub, it's a purely front-end only web app. No backend. That means their browser downloads the entire codebase instantly. The codebase is easily accessed, saved, and modified by them, if they want to. The calculator was possible over a decade ago. With some of the stuff hitting JavaScript within the last few years, you can literally build anything you want in that fashion. Games, communication apps, etc.
  - mindslight 10y ago
    > The codebase is easily accessed, saved, and modified by them, if they want to. This is harder than it already being local, or apt-get install turning into apt-get source. Likely, if they find your app interesting they will keep using it directly from your URL and end up at your mercy. And that's the crux of my point - the assumptions of the platform are at odds with software freedom. Not in a way that is incompatible - as this is computing, we are always free to ignore the properties an abstraction provides and build a completely new one over it. Your argument is analogous to a C programmer insisting that if one just follows the rules, C is just as expressive and safe as a higher level language. People are not perfectly rigorous, and the properties of the basic abstractions matter. "Web" implies executing code off of a third-party server. Yes, there is a pathological case of having your own server (and that may become more popular one day), but that does not change the term's current wider definition and the general culture around it.
    - maratd 10y ago
      > This is harder than it already being local, or apt-get install turning into apt-get source. Likely, if they find your app interesting they will keep using it directly from your URL and end up at your mercy. You have the same issues with the centralized package distribution system you're pushing? You're reliant on a 3rd party. How many apt-get the source? Saving source on a front-end web app is as simple as clicking "Save As" in the browser and you get the bonus of not having to compile anything. It'll just work. > Your argument is analogous to a C programmer insisting that if one just follows the rules, C is just as expressive and safe as a higher level language. Not quite. You're simply not seeing the levels of abstraction present in your ecosystem because you're used to them. Centralized repositories, binaries, toolchains, etc. > "Web" implies executing code off of a third-party server. If you follow this logic, your apt-get example is executing binaries off a 3rd party server. You sure you want to go there?
      - mindslight 10y ago
        Yes, I'll happily go there - it's a flaw. I'm not "pushing" apt, I used it as an example with a slightly better way of handling source. At this point, I probably should be running Gentoo/Nix/Guix (but I should also be running Qubes) but my investment in Debian has been too compelling. I'm well aware of sticky path dependence retarding progress, which is exactly why I'm nonplussed about working around the base assumptions of the "web" to make free software.