@22 @jer_ if you use Cloudflare, Google, etc. scripts though those companies are doing invasive things ... not sure what you mean otherwise since many sites which require JS are powered by some form of database and my own approach to building websites is traditional

@bookandswordblog you’re right, many web apps do need a persistence layer and therefore need a backend server to either host or proxy a database, but *tons* more sites are client-side only, either interacting with public read-only data feeds (a SQLite database, REST endpoints, etc.) or don’t even use any data (I have a tiny app that creates Ruby HTML tags for me 😅 for example for East Asian languages). Such sites tend to be built by JavaScript bundlers that combine open-source libraries like React with the app’s custom code and output a static JS file that can be served by an ultra-low-cost static provider—baked not fried (http://www.aaronsw.com/weblog/000404). If I wanted to make a no-JS version of my little Ruby HTML converter site, I’d need to have a high-cost, high-electricity compute node running Node.js or Python or C application to service each visitor’s request and *dynamically* generate programmatic output.

You’re also right that many devs will host their client-side-only sites on free static providers like GitHub Pages which are built on top of CDNs and cloud storage providers that are storing and mining visitor logs. There’s some room for shenanigans. There’s also a lot of room for privacy: I can guarantee visitors to my static site that nobody will see the data they enter into it because no packet ever leaves. And static sites offer the ultimate immunity to vendor lock-in: literally anything can serve dead bytes on disk.

Hopefully this explains why I really like shifting the computing (the “frying” per Aaron Swartz) to the client with JavaScript. I am really proud of how many great apps people have made this way. We’ve probably alienated the subset of users who disable JS by unkindly forgetting to add rich <noscript> content but this was a good reminder to avoid that discourtesy.

@jer_

@22 @jer_ there are some assumptions in that "client-side only" phrase (I would have thought it excluded all sites backed by databases, you seem to include sites backed with a database which visitors can't edit) which I am trying to unpack

@bookandswordblog @jer_ thanks for bearing with my incomplete walls of text! Read-only databases are topologically equivalent to static files (see the fantastic https://phiresky.github.io/blog/2021/hosting-sqlite-databases-on-github-pages/).

I even consider JavaScript apps that interact with public REST APIs like Wikipedia editor statistics or weather, etc., to be in the same family of client-side apps: they’re talking to servers hosting pure data with no rendering, and it wouldn’t make sense to ask a web developer who made a React frontend to one of those data feeds to move to a server to just to render HTML.

If you don’t need to “log in” and persist your own content on someone else’s server (so when you have no need to collaborate with other people, no need to save your work for later or access it from a different decide), client-side JavaScript is a very valid choice, with noscript content explaining to no-JS folks why they should enable JS, i.e., what the app does, as a courtesy.

@22 @bookandswordblog all of this that you're describing, though, is foisting off the developer's inexperience on the user. That's a bad web app. If there are elements of basic usability that don't have to be client-side that are exclusively client-side, that's bad implementation.

I'm not talking about elements that require client-side work, but websites who just show broken images without js (or nothing at all) are bad.

@jer_ hmmm. Again many thanks for spending time reading my scattered thoughts and thinking about and writing a reply—I am supremely grateful for your thoughtfulness.

I invite you to consider it less a matter of developer inexperience as much as a rational decision. Again, starting from the premise that JavaScript single-page app (SPA) libraries have revolutionized how we write web apps, consider a developer who reached for this tool to design her prototype and, after proving its business case, is finally evaluating whether a rewrite is warranted—that is, should she now move all the rendering logic to a backend language, in order to live up to the standards of a microscopic minority of users who turn off JS? We know she’s not neglecting users in developing countries—their phones tend to include quite capable JS runtimes—nor is she neglecting disabled users. Her choice of tooling resulted in an app that’s only “broken” in a narrowly-defined perspective that reminds me of something our hero @nolan lovingly called “tech veganism” in https://nolanlawson.com/2019/05/31/tech-veganism/

I don’t want to try and change your mind—I personally do respect the no-JS folks and as a courtesy to them I will spell out why they should turn on JS to appreciate a site I made in the <noscript> tag. I also love exporting websites with LaTeX equations and diagrams and source code snippets fully baked to HTML and SVG, no JS needed, to enjoy that sweet static files hosting. And finally a no-JS browser is of course in my toolkit for surviving the web (along with ad blockers and DNS blockers and proxies and VPNs and virtual machines).

But I personally can totally see why something that *could* have been a static site like Imgur or Twitter or Medium still decided to do rendering and interaction to JS. To me it’s a reasonable choice and I disagree with calling them “broken”.
@bookandswordblog

@22 @jer_ @nolan I brought it up before, but what does "web app" mean to you? To me it makes no sense to call most types of websites an app, in the way a browser mail client or GIS based map maker is an app. As I and Jer said, our main objection is not to "programs in a browser" requiring active scripts but simple displays of information requiring them.

If its all one page which loads and reloads, that suggests a site is more a web app than a web page.