I kind of wonder why it took this long for someone to try this approach. It just makes a whole lot more sense on the surface.
Interesting thing I just thought about: since users don't get to choose their system webview, I wonder if this has the potential to broaden browser diversity a little bit?
A major benefit of Electron is that you can develop against a single browser and runtime version and don't have to deal with all the small but time consuming compatibility quirks.
On Windows the default webview engine is still not the new Chromium but old Edge (WebView2 has to be installed separately), Safari has a fair amount of issues and gtk-webkit can be very quirky as well.
This is not so significant if your app has a web version and you have to deal with it anyway , but thereare a fair amount of Electron only apps. Or even with a web version, the desktop app often has special features not present on the web.
But webviews are as much a moving target as browsers and often behave differently from the regular browser version in a myriad of ways, so you need an even bigger testing matrix.
Webviews are also hard to properly secure and don't provide the extensive APIs of Electron, which you often have to bake yourself in a wrapper written in, eg, Rust or C++.
I say all this as someone who hates how resource hungry Electron apps are and groans each time about yet another one on my machine, but I understand why it is so popular.
> don't have to deal with all the small but time consuming compatibility quirks
> Webviews are also a moving target so you need a big testing matrix
I say this as a front-end web app dev who would most likely be affected by this kind of shift: it might be slightly less convenient, but I wholeheartedly think it's worth doing, on several levels.
People built web apps 15 years ago in the dumpster-fire that was web standards at the time (and frankly, web tooling too). Browsers, even the good ones, didn't update themselves automatically. Most of the world was still on IE which was actively hostile to web standards for the sake of lock-in (and users weren't even on the same version of IE!). There was nothing like Babel to smooth over the rough edges, there were no polyfills. And even the standards that existed, when you got the opportunity to use them, mostly sucked.
I think, given all that, we can deal with supporting recent-ish Chromium and the latest Safari, if it means bolstering web diversity and saving an order of magnitude in RAM and storage.
Also: a lot of our "Electron" apps are essentially Electron frontends to the existing web frontend.
For example, Slack-the-Electron-app and Slack-the-web-app are fundamentally identical, aside from likely some platform integration features and so on. That means that, likely, a lot of these Electron apps would already work in, for example, a Safari or Edge webview, since the web version has to anyway. I sincerely doubt that they're taking the time to strip out that compatibility, since the only reason to do so would be efficiency and if you're using Electron that's not a priority.
IOW, for many applications, there's a near-100% chance that everything would function with little to no extra work, and for most of the rest it would likely take very little.
Slack originally was platform specific webviews, and migrated to avoid issues like what is mentioned in this thread. One of the engineers who did it has even commented on HN to explain this.
This is probably the comment you're talking about: https://news.ycombinator.com/item?id=18763449. I was unconvinced then, and I am even more unconvinced now as I run a fully native Cocoa Slack client on my Mac that was written by one person reverse engineering APIs.
Nobody ever said that Slack couldn't be implemented in Cocoa, and it's really not remarkable that someone did it. ;P
You and I both know that one person can write UI in a vacuum, but there's a litany of other challenges that come up when positioning an application across platforms - sharing code, resources, etc. If all you look at it on is a technical level, you're doing the problem a disservice.
That same reasoning is why Google is having a YouTube app on AppleTV which does not use native widgets, so bluetooth keyboards don't work, scrolling is off etc.
It's not that they cannot afford building 10 native AppleTV apps in parallel and pick the best one.
This sounds like you didn't look up and read the comment in question. Problems with cross-platform webviews stopped being the HTML/CSS/JS layer long ago; it's integration, availability, and rate-of-updates that are common blockers.
Call me when Microsoft actually ships WebView2 with the OS.
"This is literally what Slack moved from years ago, because it was very problematic, let me tell you why since I'm the one who did it.
* Adding new APIs was a huge pain in the ass, you had to write all of this ugly bridge code in Objective C, and the APIs ended up being super unnatural on the JS side. As a result, desktop integration wasn't done much since it felt like a "Black Magic" type thing. Writing new APIs in Electron is Just JavaScript, so you can make something much more natural.
* We could literally do fuck-all for people running old versions of macOS - you can't upgrade Safari on old machines, you just Get What You Get. For awhile, every YouTube video on older macOS versions had a pink shade to it. Users write in, "This is clearly wrong!" "Yep, but we can't fix it. Sorry."
* And big spoiler - WKWebView uses _basically_ the same amount of memory give-or-take a few MBs as Chromium. It's the content that sets the memory usage bar, not the host. "
> That means that, likely, a lot of these Electron apps would already work in, for example, a Safari or Edge webview, since the web version has to anyway.
Coincidentally, Slack doesn't fully work in Safari or traditional Edge. Video calls only work in Chromium-based browsers, and in the Electron app which is of course also Chromium.
This is actually the thing that finally got me to switch back to Chromium last week, after using Safari and Firefox for the better part of a decade. The Electron app sucks, I don't like switching between browsers for usability reasons, and I need to make Slack calls for work.
That's fine for someone who already keeps current with front-end stuff, but someone who is a desktop app developer is more likely to just go native than try to make a desktop app that tracks 3+ browser engines of wildly varying versions (luckily 7 is out of support now, but it still might be desired to support for some desktop apps in which case now you have IE to deal with). Electron gives the promise of total code re-use across platforms which is what makes web technologies more suitable for desktop apps.
> Electron gives the promise of total code re-use across platforms which is what makes web technologies more suitable for desktop apps.
I am not sure what the point is of Electron: a shitty-resource hungry app that literally is a wrapped web-app. Instead of writing platform specific apps that leverage the power and capabilities of those platforms, we get a jack of all trades and master of none. Electron is a business decision, not an enlightened technical one.
What value does electron add to the user? I would argue none. That Slack can’t find the resources to write actual Swift for a Mac App is just amazing to me. Instead we are essentially interacting with a lowest-common-denominator web app. Is a chat app that hard to write natively? We’d get better performance, a smaller footprint, and a more polished, platform-specific product.
> It adds a lot of value to Linux and to a lesser extent Mac users: before a electron a lot of apps didn't have a version for these platform at all
What value does _this_ add over a web app?
I've been an exclusive Linux user since college so I'm aware of the disadvantages of app/service unavailability (I remember the dark days of running Netflix in a windows VM that I barely had enough RAM to run...).
But what's the advantage of using the Slack app over a browser tab? This is a sincere question, since enough people reference the benefit of faux-native wrapped webapps that I assume there's some value.
I seem to be the only one in my company who doesn't use Slack's app, and I'm always shocked when I see a colleague's computer freeze and have them say "ugh it's almost definitely the Slack app". I work in autonomous vehicle engineering, so these are _monster_ machines I'm talking about, capable of running the entire car's stack locally. The only advantage I can think of are desktop notifications, but 1) I'm pretty sure browsers support those and 2) is that really worth using an app that's amateurish enough that it freezes your computer regularly?
I'm on Ubuntu and use the Slack app, which never causes any freezes or anything of the like. The advantages is that i know where it is ( in my taskbar and launcher) compared to it being one of my 500 open tabs, and the taskbar icon shows if i have unread notifications or messages. Furthermore, i can use it with multiple workspaces at the same time.
Ah yes, the panel icon. This makes a lot of sense, thank you! My os is _heavily_ customized (i3 + wmctrl + heavy scripting), so I forget sometimes what the standard floating-window workflow looks like and how difficult it can make it to manage and quickly access important windows/tabs.
> It adds a lot of value to Linux and to a lesser extent Mac users: before a electron a lot of apps didn't have a version for these platform at all.
I prefer web versions of those apps on Linux than electron.
Sorry, I want to run just one browser and it is not Chrome based one.
And for sure Electron apps don't allow that ease of cross platform development as one thinks, I would argue that it is on the same level as a native app with good abstraction library (see Telegram client source on github).
It is more that they don't want to hire desktop developers (or just ones that want to learn something else than JS).
For an app with no source available publicly, I would much rather have a web app in a sandbox than a "native" app (whether native or electron-style) that I would have to trust.
Electron exists because there's a horde of developers whose only tool is a JavaScript hammer, so every problem looks like a nail. That's how you end up rewriting `dd` for Chrome[π].
I think we got there also by desperation of lacking good cross-platform GUI toolkits.
The current state is, you either use something that looks like Windows 95 in all platforms, or web app.
What the web made possible is to anyone to create a decent custom UI with little effort. I still remember how messed up was to create a custom component for Java Swing, which I believe is still one of the easiest to do. How would you do that for QT? GTK? WxWidgets? etc
Also totally inflexible, you better have the right-sized screen/resolution: too small and you can't see that "ok" button; too large and you'll have a ridiculously small and crammed winform to fill/see your information.
Interesting Telegram has both a native (Swift/Obj-C not sure which) and an cross-platform-codebase app for macOS. The former is called Telegram for macOS and the latter Telegram Desktop. Both are on the app store even. So it would actually be a very good case to compare performance, etc. The native one presumably shares a lot of code with iOS which is probably why it exists at all, and it supports secret chats. Otherwise they have similar but slightly different design and features.
However, I think the Telegram Desktop client is not Electron either, it is just regular C++ compiled on different systems.
I might be crazy, but for me it's faster and more convenient to download and run Etcher than to read up on dd command arguments and device paths. Storage space is plentiful for me. RAM isn't an issue as it's only run once in a blue moon.
Is it crazy bloated and resource hungry for what it does? Undeniably! Does it matter? Not for me.
Electron resource use can be painful for apps that are running constantly, like Discord, VSCode and Slack. For utilities like Etcher, who cares.
I realize I'm replying six days later - but I can't help mentioning that the "dd situation" is even crazier than it first appears: you can just use `cp foo.iso /dev/sdb` or `cat > /dev/sdb`.
The actual duplication is done in the Node.js runtime, not the V8 runtime inside of the Chromium GUI layer.
I do wish they had a command-line version that just ran node without Chromium, but there is a use-case to have a simple foolproof application like this; these days there are a lot of people who need to image drives like this and would be best not to mess with dd on their first time.
And dd doesn't exist for Windows so you've got to use some tool or another like this.
Honestly, this site is the only place I really hear about the supposed horror of electron. The resource issue doesn't seem to effect me or people in the 7 slacks I'm in all day.
My 6 year old MacBook runs slack, Xcode, android studio, safari, chrome, messages, vs code, vim, iterm2, preview, battle.net and hearthstone all the time without issue, really.
I think for the average user, they'd rather have more features across platforms more quickly than a smaller resident memory footprint and install size.
For a counter-anecdote, I've seen the Slack app freeze up colleagues' workstations many times, and these are monster machines designed to run an autonomous vehicle stack on raw sensor data.
Electron app is not a native Linux client, it is just a browser app.
If you ran that app in a tab it gives you much better performance and less memory used.
That all depends on how the app is written. The idea of Electron is to bring the web app frontend experience to apps which are fundamentally not able to run in a browser, such as VS Code, Etcher, etc. Writing to raw device files, listening to ports, using the full CPU resources and timers not available in the browser, using more memory, etc.
If you just package an existing website into an Electron app that is a waste of time and user resources. Electron is intended to combine a web-based presentation/UI layer with a Node.js layer that supports access to native resources, unlimited storage, persistence outside the browser, etc. If you are not making meaningful use of these capabilities a PWA would be sufficient.
Overall, I do prefer native apps, although there are examples out there such as VS Code that have achieved UI excellence without it. Although, I really only like that on macOS... on Windows I prefer the older Windows Forms style rather than the new Modern UI, and on Linux there is a bit of style mismash between Qt and GTK apps already.
> What value does electron add to the user? I would argue none. That Slack can’t find the resources to write actual Swift for a Mac App is just amazing to me. Instead we are essentially interacting with a lowest-common-denominator web app. Is a chat app that hard to write natively? We’d get better performance, a smaller footprint, and a more polished, platform-specific product.
Exactly, one can look at Telegram app, it is clean, polished and open source client written as a native app that works across many platforms (and one of few that has Linux client).
Electron should be banned for the environments sake if the other reasons doesn't persuade, it's inefficient computation to energy ratio using Electron apps, when there is a lot of users like with VSCode or Slack it probably affects the global climate in a negative way
I see a lot of people in the tech industry deeply concerned with human impact on the environment and our diminishing prospects of repairing it without drastically altering our lifestyles.
At the same time the general consensus in the tech industry towards the steady increase in power requirements for everyday computing seems to be indifference.
These two viewpoints cannot be reconciled.
Every month when this Electron discussion comes up there are so many who dismiss the egregious resource usage outright on the basis that it's trivial for modern computers. I'm one of those people who use Thinkpads in 2021 for my personal computing and have no problems with it. I will never buy a Mac, and I don't want to unnecessarily spend hundreds of dollars on a modern system when refurbished ones from a decade ago work so well. I avoid Electron apps on the basis that they flatten my battery quickly. I switched from using vscode to vim on my x200 and my battery life tripled, I'm not even exaggerating here either.
No, they are completely reconcilable. The amount of power it takes to run a powerful consumer CPU is trivial compared to the amount of power used in many other common applications like heating, vehicles, etc. This effect is particularly pronounced when you consider that laptops use way less energy than desktops (on average) and mobile computing as a whole is displacing desktops outside of niches like PC gaming.
The environmental issue doesn't make sense IMO, but increasing hardware requirements does have real impact on accessibility.
These effects scale. If every single laptop uses a small amount more power than is necessary, the relatively minor increase in energy use for one computer totals up to a significant increase in overall energy usage. Not to mention the unnecessary obsolescence of hardware. The externalities of manufacturing new hardware - and disposing of old hardware - are extremely significant. Just so what? So my perfectly good 10 year old laptop has enough RAM to run 6 Electron apps at once?
Sure, your ceiling lights use less energy than your fridge. That's not a justification for everyone to completely ignore turning them off when not needed.
If electricity consumption of all computers in the world doubled because of Electron apps, the total annual electricity consumption would increase by 0.01%. So probably false.
In 2012 PCs (only PCs, not network appliances, not data centers, not phones...) were estimated at 1.5% of global electricity use, so you're off by a good margin. And that is deeefinitely not negligible, nor is the need (and eco impact) to buy newer computers because a 2016 MacBook Pro is unusable when your work requires 4/5 electron apps side by side.
No joke, if it weren't for the resource usage of vscode and the couple of other Electron apps I rely on, I wouldn't consider upgrading my workstation (which is already new enough to be energy efficient enough but maxes out at 24GB RAM) but here we are.
Maybe not banned but they could do a carbon tax like they do with motor vehicles. Or maybe like with appliances where you have an energy rating, with stickers telling you how much electricity it will use per year, possibly with tax incentives to pick the most efficient.
Problem is: how to test that? Even with the most straightforward tests, there are cheaters. Appliances have a special "eco" mode to meet the specs that most people won't use because it barely does the job. And for cars, well, VW have shown us how far companies are ready to go.
I run three browsers on my desktop, because the number of things that don't work properly in anything except recent Chrome is enough that I keep encountering issues, I don't want to use Chrome for my regular browsing, and both Safari and Firefox have issues on some sites.
On my tablet, a number of current websites don't run even on Chrome, because the tablet isn't particularly new, so it doesn't have current Chrome. It's still great hardware in perfect condition though, everything that does run on it is perfectly smooth and so on, and I still use it. Youtube in particular is fine. Literally the only reason to buy a new one would be to satisfy the software treadmill.
My old phone didn't run Chrome (or Safari). It had Firefox. The only reason I switched to a newer phone, using Android, was because of sites that gradually didn't run well on Firefox any more. That was annoying, having to ditch a perfectly great phone and replace it with something better for the web but worse for other things.
I have a Macbook Pro at home. I'm pretty happy with it, it's still a fine machine, I use it for work every day. It runs the latest Safari that's safe to run on it, but that's not the latest Safari, because it would have problems running the most recent OS.
This brave new world of "I'm sure we'll be fine supporting recentish Chrome and latest Safari" is a world of exclusion. I don't like the memory usage of Electron, but I prefer it to applications that are a bit broken!
Also, to be honest, I've been looking at memory usage. Electron uses much less than any of the three big browsers. You may know that Chrome famously starts a new process for each tab, and Safari does too. Firefox isn't far behind. As a result, the memory saved by running a new page in an already running browser compared with Electron isn't as much as it was long ago, back in the days when tabs shared memory. This is visible with Safari, as the website name shows up in the system utility for viewing process memory.
I know, that feels wrong somehow, but it's true.
Finally, when using a system WebView, will that save memory if it's opening a separate instance of the system's browser? It must be separate in some ways, if opening an application that uses a WebView doesn't open all your saved tabs from last time the browser was open. (And it's a horrible experience if opening an application does open all the other tabs.)
Running tabs in a separate process has been a boon IMO. It’s better security and a crashed tab doesn't bring the browser down and an unresponsive tab can be killed independently.
I agree, however that is part of the argument for why Electron RAM usage should be compared against browser tabs, instead of assuming that Electron must be using a lot more ust on principle because it's separate.
One major benefit of running webapps in an actual browser like Firefox, instead of Electron, is that I can have uBlock and a handful of other defensive plugins, to prevent shady things happening behind my back.
but, conversely, browsers tend to have access to a lot of ident/auth goods that a standalone electron wouldn't, even if you're using containers. also it's perhaps harder to phish someone running an electron app because outside webpages open a real browser.
Did you try Brave? I'm yet to encounter anything that works in Chrome but not in Brave (save for the occasional site that breaks because of ad/trackblocking, which can easily be disabled)
> Most of the world was still on IE which was actively hostile to web standards for the sake of lock-in (and users weren't even on the same version of IE!)
More like: A lot of corporate users mandated a specific version of IE and wouldn't allow anything else to be installed.
Some corporations did that. Most didn't. The main reason IE dominated was that most people didn't care, and stuck with the default. The main reason old versions of browsers were more common than new ones was that self-updating browsers hadn't become a thing yet, and most people didn't care, and just stuck with the default. The reason most browsers stay up-to-date now is that modern browsers are all configured to self-update by default, and most people don't care, and stick with the default.
I have never bought a prebuilt Windows PC, so this is pure speculation, but I am wildly guessing that the dominance of Chrome is because computers are now coming with Chrome preinstalled and configured as the default, and most people don't care, and just stick with the default.
My experience was that IE finally got bad enough (around version 8), and Chrome got good enough, that even "grandmothers" started having everyone telling them to use Chrome, and it was so incredibly superior that you didn't have to be technical to appreciate it.
Now the reason everyone uses it is because it's psychologically the new "default", whether it is on the actual system or not. Everybody knows that's just what you do when you get a new computer: you go install Chrome. And the irony is that even now, with Edge being basically the same thing, most people don't bother to re-evaluate that default assumption that "the windows browser is bad". They just use Chrome.
> I have never bought a prebuilt Windows PC, so this is pure speculation, but I am wildly guessing that the dominance of Chrome is because computers are now coming with Chrome preinstalled and configured as the default
I'm not aware of any Windows machines shipping with Chrome pre-installed. But as soon as you use any website that's owned by Google you get spammed with install links for Chrome.
Entire generation where? Let's say I saw exatly 1 guy with a chromebook through my life, and likely chromebooks weren't even officially sold in my country. And still Chrome is the most popular browser here.
Chromebooks outsold Macs in 2020 and schools all over the US are pushing Chromebooks on students because they're affordable and accessible to most people.
I remember there was a long period where everything would bundle Chrome. One wrong checkbox and your computer would be infested. Adobe Reader was one prominent example.
In 2008, when IE7 came out, I cried of joy. I turned 22 that year, so the 7 years with only IE6 and Firefox was a significant chunk of my life. Firefox showed us that the web wasn't finished and could still be improved, and then came Firebug which made Firefox first web dev the obvious choice. But the vast majority still used IE6, so there was always a miserable step in the project of fixing all the IE bugs. After doing it for years and knowing your foe, it was by then a relatively quick and manageable experience. The real pain was in knowing all the CSS2 features and other goodies that could never be used. Seeing Microsoft recognizing that, yes indeed, the web wasn't finished in 2001 was actually an emotional moment at that point.
So yeah, the situation is without doubt much better now. I think people tend to forget (or aren't aware) that Microsoft actively said "there's no more work to be done here" and stopped all progress for 7 years. There's been some progress on the web since 2014, as a point of comparison.
I've had to say this a fair few times here, but I'll keep saying it.
IEDevToolbar, which Firebug was pretty much a clone of, came out a year before Firebug. And it took Firebug years to catch up in feature parity, as far as I remember.
So, firebug didn't make it the dev tool of choice. We all just hated microsoft. And tabbed browsing was great.
I was young and dogmatic, so I did admittedly despise Microsoft back then. I can't recall I ever heard of IEDevToolbar, though. Did it really feature the DOM tree view where you could do CSS manipulation? I always had the impression that it was a Firebug invention, but I could be wrong about that.
I worked on a highly complex web app for ~3 years, and we made a point to support Firefox, so whenever somebody had a Firefox-specific bug we wouldn't tell them to just use Chrome, we'd fix it. We only had 2-3 Firefox-specific issues in that entire time.
I have heard that mobile safari can be a bit more unruly than the rest, though it isn't really relevant to the question at hand
Similar. Built a very complex app officially aimed at oldIE but made a point of ensuring it worked on other browsers too, largely on a skunkworks basis. A few years later (when we're no longer actively working on the app) everyone switched to Chrome and it Just Worked. Someone tried it on an iPad and it Just Worked (modulo some fat-finger challenges with small UI affordances).
At the time, jQuery was a big help in this, especially in hiding the event model mess. Things are a lot better now. Although I do wish we had a strict compatibility-checking layer that'd warn you if you used any feature of the web platform that wasn't supported by everything you wanted to support.
Anecdotally I wrote some JavaScript just today (I'm not a frontend dev but I write some simple stuff for fun sometimes). It seemed to me that every modern feature you'd care about is now widely supported. I used the JS fetch API and template strings and these things just work in modern browsers. I even encoded my web fonts in WOFF2 only, because it's so widely supported that I didn't see the point in bothering with WOFF1 too.
> Webviews are also a moving target so you need a big testing matrix.
Even worse, an OS update after you ship your app can spontaneously break it on end user's machines without you knowing or being able to do anything about it.
At least with the web, when new browser versions come out, you can fix your site since users essentially "install" it every time they refresh the page. But an installed app that runs on top of a spontaneously updated framework is the worst of both worlds.
Seeing as the auto-updater wouldn't have anything to do with the webview, I don't see how that's any more likely to happen in this case than it would be in Electron, or any other desktop application for that matter
But why would the system webview be less resource hungry? If it runs in an own process instance, the only memory you safe is (maybe) the executable code (i.e. maybe 100MB max).
Or you would need some common runtime which is shared across the OS, and somehow use IPC to it. Then you can potentially save some more memory. But I don't think this exists. And I'm also not sure how much you really would save.
The advantage of using the system webview is that you can update this component, in case there are any vulnerabilities or so.
100 MB / app is significant, especially on laptops.
Slack used to use so much memory on Linux that I couldn't afford to run the desktop app and ran it inside a Firefox tab for years... seems to have gotten better somehow.
> don't have to deal with all the small but time consuming compatibility quirks.
Good, then developers might start to develop against standards instead of handpicked browsers. Or if they use a future (unstable) standard that is their own headache (as opposed to user’s; which are subsequently in practice forced to use chromium derived browser).
An app developed against a single browser (as opposed to the web standard) should be considered what it is, an ugly hack, and developers that do it should only do it for their own private apps never to be published until it is standard compliant.
This is about desktop and not web development though. I agree that pages available on the web should be standard compliant and work with all browsers.
On the desktop you are developing against a runtime that the developer of the app gets to choose, so why not choose one that has less potential for quirks?
Because then you'll get a more efficient app that people don't constantly complain about?
Especially on macOS, where Chrome/Chromium have historically been extremely inefficient, swapping the Chromium engine for WebKit2 WebViews would mean battery, memory, and CPU savings which would absolutely be worth the effort IMHO.
Apart from the install size which is still not optimal, it is very much possible to write an efficient Electron app, as evidenced by Discord and Element.
That Slack, the poster child of bad Electron apps performs so badly must therefore have different reasons. Exactly the same way you can write websites in a way that is more or less efficient you can write Electron apps of varying degrees of efficiency.
Especially on MacOS where Webkit usually behave differently? Like, why would webkit still require webkit prefix for Web Audio API, which is like 8 years old at this point.
To me, develop for Electron is more fun than normal web because I don't have to check whether the new web feature I am using is available everywhere (especially Safari) yet.
Yes, it doesn't yet render pages as good as other browsers, but we can tweak our code for it. 10x better performance and 100x less memory usage is a win.
One approach that you can take to deal with this is to mentally back way the fuck off the 1990s on what your UI expectations are.
Give up pixel-accuracy and fine-grained control of your app's UI layer, and just treat it how people like Peter Norvig and Chris Lattner treat their home pages.
There isn't any browser with even 0.1% share that can't render those things.
You also don't really have to go as far as those guys do, leaving the font 14-point Times New Roman and using HTML 1 tables. You can add some style, fonts, etc... just don't add anything that would make it _not work_ if somebody opened it in Netscape Navigator 4.
There is definitely a big downside, in that your app won't look all dope and modern. For many apps that is a deal breaker. Also, some UIs really do require a higher fidelity UI. Not all UIs are just like, buttons and menus (although a lot are).
But there is also a big upside: durability, in the sense that your app's UI will work on most 10-year old computers, and most computers 10 years from now.
Yes, but it is not installed by default. You have to install it separately, at least for now,which makes things a lot more complicated if you want your app to rely on it.
Edge is pushed to the vast majority of PCs, and on those machines, I believe it provides a system-level webview2 component. You can choose to package your own webview2 into your app if you want the extra compatability that offers at expense of app size and ram usage
I wonder if it would make sense for Tauri to support shipping a bundled webview engine for platforms where the native webview is outdated or otherwise problematic for your application? With the idea that you could eventually migrate to system webviews.
That way, you're building with cross-platform in mind but still have the option to pick a stable target if you need it, compared to solutions like Electron where you don't have that choice.
>> I wonder if it would make sense for Tauri to support shipping a bundled webview engine for platforms where the native webview is outdated
I guess you can ship an electron version for these platforms where native webview sucks. Chances are that once you do the electron app you start doubting the benefits of shipping a webview based app.
> A major benefit of Electron is that you can develop against a single browser version and don't have to deal with all the small but time consuming compatibility quirks.
Somehow I doubt the ideal solution is to bundle a web browser with every app. Electron feels like a prototype that went out of control.
I think they're overstating the browser compatibility problem too. It was really bad 15 years ago, but today it is almost a non-issue. Between significantly improved standardization of web technologies, and new tricks like polyfills, I think it has become pretty uncommon for apps to not just work on multiple browsers without any tinkering.
"almost a non-issue" doesn't sound very convincing to me.
As someone who doesn't write GUIs often, I just don't want to have to worry about it at all. What's the cost of electron over webviews? 100mb hdd space? On the other side of the trade, never having to worry about cross platform concerns again? Not having to support or test on multiple platforms?
Seems like a no brainer to me. But it's entirely possible, even likely, that I'm putting too much faith in electron, or that there are better options I just don't know about.
I really don’t share your point of view, but I understand it, for me feels like we as developers are getting worse and worse with the time, we get every year more powerful machines but our software is getting worse every year, who cares about performance if is cheap to develop, maybe if our apps uses less resources it would have less impact in the environment (less energy consumption is better for everyone, except for the lazy developers).
I really hate electron apps, for example you have teams, that is written in electron, in linux is somehow better than the web version but is just way worse than the windows/mac versions, in web you can view 2 cams, linux 4 but in mac and windows you have more cams at the same time and also a lot of features that are missing in linux, so I don’t see too much the benefits of electron here, just pure laziness of microsoft not willing to invest in a better cross platform app.
I also hate electron apps, I refuse to install them on my personal machine, and use a(n independent) web browser.
Efficient software is great, the cool parts of the projects I work on are efficient. In my case, the GUI doesn't need to scale in the same way, and I'm pretty confident the customer would rather have better cross platform support (at a lower cost than if we had to test a bunch of OS web views) than 100MB of their disk back.
But yeah, there are plenty of examples of software I use where I disagree. Although this is often based on my aesthetic dislike for an inelegant system, than on any kind of resource constraint.
> I just don't want to have to worry about it at all
And in doing so you're prioritizing your needs over all of your users'. I think it's fair to claim that a minor convenience to you, at the cost of inconvenience to all your users, might be something you may want to deal with.
I cordially dislike Electron apps, but once one factors in opportunity cost given limited developer manpower, the trade-off you're choosing to make here is quite likely entirely rational.
I would prefer to live in a world where I didn't believe the last sentence to be true, but in the current world I still think it is :)
In my experience, unfortunately it isn't true that apps work with very high reliability. This is why I use multiple browsers. I don't want to use Chrome, and I encounter new websites with functionality that don't work properly in Safari or Firefox quite regularly. Some of this is even quite basic functionality.
From things like shopping carts where the final payment button doesn't work in Firefox (nothing happens), to video conferencing that sort-of works in Safari but then the audio breaks while it is reliable in Chrome.
Even GitHub doesn't render properly in Safari occasionally (the page is full of weird size text and giant rectangles). But that's almost certainly a Safari bug.
This is exactly the problem I was suggesting might be improved by an increase in browser diversity. 9/10 times, the issue isn't the browser itself, the issue is that the dev didn't bother to test and find the (handful of often small and easy to fix) bugs in other browsers. If they were forced to do QA - and ideally, had to adhere better to the actual standards to make their own jobs easier - you might have an easier time using those apps (and the libraries they depend on) in other browsers.
I think this is more a problem with developers working in Chrome and using Blink-specific features. I stopped having compatibility issues with my apps when I moved to Firefox years ago.
Some of it is Blink-specific, but some of it is silly Javascript stuff. How else do you explain a shopping cart final payment button not working.
I like development in Firefox too. When I was doing it intensively (just ~2 years ago), I found some of the CSS I was using behaved differently between Gecko, WebKit and Blink, though (and differently again in Android's pre-chrome browser), so I reconfirmed that I do need to test some things across all browsers. By differently, I don't mean prefixed browser-specific CSS (although there is that), I mean standardish things like table row height calculations showed up significantly differently, certain border highlights, and a number of WebKit rendering bugs that Gecko did not have. Key-event based text editing and wheel events are also significantly different, even more so when you include the common mobile browsers.
At least with writing uncomplicated web applications it was still quite easy to have it compatible with the major browsers, you just had to test the web application with them as you wrote it to make sure. Compatibility problems didn't even come up very much while building it if you stayed reasonably within the boundaries of standards that were well entrenched. The biggest issue was how little CSS IE6 supported, if you did anything fancy with CSS and it didn't work with IE6 you had to include an IE6 specific hack, fortunately that was easy and worked completely fine with other browsers.
There were problems with internal off the shelf applications whose user interface was built for IE6 only. The funny thing is those applications still worked well with the Firefox of the time, but IE9 broke compatibility with them and the applications didn't work on IE9, but IE10 got released and was compatible with them again.
People have been trying the WebView approach for years. (Projects include: Quark, Electrino, DeskGap, Revery, and Neutralino.)
Historically, it worked terribly on Windows.
On Windows, if you used the OS WebView, your Windows 7 users would be forced to use IE11 to run your app, even if they had a newer/better browser installed locally. On Windows 10, you'd get whatever random version of Edge was installed, or the new Chromium-based Edge.
In 2021, we're in a new era for Windows WebViews, thanks to Microsoft shipping WebView2 at the end of 2020, which ensures that the OS-provided WebView will be a modern version of Chromium. https://docs.microsoft.com/en-us/microsoft-edge/webview2/ Tauri supports WebView2 and I bet it will work a lot better than historical approaches.
Support is still pretty dicey on other platforms, though. macOS doesn't have anything like WebView2, so if you want to support a five-year-old version of macOS (which isn't that old in desktop terms), you'll be forced to support and test your app on a five-year-old version of Safari. (The user might have upgraded to a two-year-old version of Safari, but they might not, if they prefer Firefox or Chrome, and that's now your problem.)
The easiest and best way to improve the user's experience on old OS versions is to provide your own copy of Chromium, like Electron does.
At that point, if you've shipped an Electron app for macOS and Linux, maybe you just wanna ship an Electron app for Windows and call it a day?
Having said that, if you can keep your OS-version support matrix tight, Tauri might work OK for you.
Fortunately MacOS users upgrade very quickly. High Serria which is slowly waning at 7% came out in 2017. Compared to windows where twice as many (16%) at still using Windows 7 and won't be able to download the Webview2 update which Microsoft kindly provides.
We use the WebView approach for Kreya[1], though not with Tauri. We had some occassional issues with Chromium/Safari differences, but otherwise it works fine.
We use WebView2 on Windows, Safari on macOS and WebKit2GTK on Linux. Not shipping a whole Chromium installation on each install and update reduces the installer and on-disk size a lot!
> you may install Kreya by downloading the tarball. Note that both libgtk-3 and libwebkit2gtk-4.0 dependencies are required. Install them manually if they aren't present on your system.
libwebkit2gtk is, itself, _an embedded browser_, 44MB installed size, plus dependencies.
It's convenient that you can make Snapcraft install it for you, but that's not really different from making Snapcraft install Electron and running that.
The Apple Silicon (M1) issue is because .NET (which we use for our "backend") doesn't support it yet.
Linux does not have a core WebView out of the box, so libwebkit2gtk is the closest alternative. Not optimal, but it works. Looks like Tauri uses libwebkit2gtk too.
Electrino (4 years old) was an experiment where they forked Electron and removed Chromium to replace it with the native OS Web views. https://github.com/pojala/electrino
Webviews don't work because they solve very different problems.
System webviews will always have different features, break things, and are entirely uncontrollable by the software distributor. This means your software will not be resilient to the future.
You cannot access the JS engine the same as you can V8 in most webviews. This means you can't interop with C libraries without some performance destroying hack like copying everything into the JS space through sockets.
This also means something as simple as reading a file has to, again, cross the sandbox in an inefficient way. You'd have the browser, it's javascript engine, and your own runtime in node or python or however too.
Electron develops patches that make chromium more performant and feature ready for desktop.
Electron conjoins the chromium and node v8 event loop to take advantage of libuv - again a performance improvement.
The decision of webviews and electron is based on what you're trying to do. If it's simple html/JS that's pretty much self contained then sure.
It's been a minute since I attempted it on Windows, but transparency of borders and borderless windows with the webview on Windows was next to impossible. It's one of the reasons we used Electron. I'm curious to know if they've solved that, but can't find my mention of it.
Yes. Why not? HTML/CSS great (from the development POV) for UI, and I can always write the performance critical code in C and linked them to the electron HTML/CSS frontend.
Presumably, if copying some data between JavaScript and native code in a slightly roundabout way was an issue, you wouldn't run things through web technology in the first place…
>I kind of wonder why it took this long for someone to try this approach. It just makes a whole lot more sense on the surface.
The current generation of web developers who grew up within the Chrome ecosystem has zero interest in spending time on cross browser support.
Even if the user had to install 4GB of chrome runtime they'd still argue it's better than them making it work in Chrome, Edge-Chrome and Safari WebKit.
Not saying everyone, maybe my companies hiring process is screwed but every developerbar 1 that I've worked with over the past 4 years has been the same and even getting them to test in one other browser than Chrome is honestly like pulling teeth.
Infuriates me as someone who's first job was CSS+HTML supporting IE 5.5-7.0, Firefox, Opera and Safari.
The amount of times I've heard "Can't we just tell them to use Chrome?" when talking about users is getting absurd.
The lack of understanding that a few hours of their job being slightly more difficult can add up to saving literal lifetimes of time saved or entire data centres of disk space when you extrapolate the savings across your user count. I know its maybe cheap to roll out a Steve Jobs quote but this one still resonates with me.
"Well, let's say you can shave 10 seconds off of the boot time. Multiply that by five million users and thats 50 million seconds, every single day. Over a year, that's probably dozens of lifetimes. So if you make it boot ten seconds faster, you've saved a dozen lives. That's really worth it, don't you think?"
Pity they can’t use Servo as the rendering engine, even if it only deals with a small subset of style/layout properties. That kind of parallel layout engine should make building fluid 60fps interfaces a lot easier.
Servo is quite large and extremely difficult to build from source. We've tried to make it work with Tauri but decided it's not worth it, at least for now.
Do you have more specifics, and/or have you advised them of the difficulties? It seems to use the same "mach" build process as does firefox, and I build FF developer edition regularly (not every day, but damn near): https://github.com/servo/servo#normal-build
While investigating this, I have deep sympathies for whoever has to work with that taskcluster silliness because yikes that is some A++ grade obfuscation as compared to a .gitlab-ci.yml or .circleci or even the .travis.yml they migrated off of
AFAIK there are no V8 <-> Servo bindings. So in theory (I don’t know Servo well at all) you could render a static HTML document but you’d have no DOM JS APIs to work with.
> I kind of wonder why it took this long for someone to try this approach. It just makes a whole lot more sense on the surface.
As others have mentioned, this is not the first time someone has tried. As for why people might not go for it:
- Electron is Chrome. Webview is: Cocoa/WebKit, gtk-webkit2, and Edge (or is that Edge and Edge/Chrome). Yes standards have progressed a ton but there's still inconsistencies and it's nice not having to worry about them.
- Electron is JS, so your app is 100% in a single language. In Webview. With these alternatives you now have to use 2 languages. Not the end of the world, but again, nice not having to deal with.
...has a fairly long history. It's great for extremely small application packages (when I tinkered around with it on macOS a few years ago I brought it down to 26 kilobytes).
The usual argument against the idea is that Electron gives you a fixed Chromium version to work against, while the system webviews are a moving target with different underlying browser engines.
Interesting. So according to @baxrob, Tauri took over control of https://github.com/webview/webview project for the past year. And now I'm told they are moving away from it.
I'm left wondering why take over a project if they're going to ditch it. For lack of a better term.
We didn't take over control, rather we helped setup an independent org around webview and other related repos. At the time, the original author of webview expressed plans to work on it a lot. However, this didn't really happen. Webview is stuck with some nasty bugs and missing features, and none of the members of the Tauri team had enough C experience to fix it efficiently. Instead, we created our own pure Rust solution (https://github.com/tauri-apps/wry). We've already given it way more features than the original webview project, and it doesn't carry the bugs that plagued webview either. The next release of Tauri (about a month or so out) will use Wry, and will have features such as multi-window and fancy window styling (frameless, fullscreen, custom controls, etc...).
One significant difference regarding webviews is Neutralino on Windows is using an outdated WebControl (MSHTML/Trident based) and this one seems to be using at least WebView (EdgeHTML based) or even WebView2 (Edge-Chromium) if available, both of which are a major improvement.
the interesting history of these sorts of frameworks is that Google actually created a framework that did this and stopped development on it. the code is still on GitHub. And there's a bunch of other frameworks that use a variety of different languages not just rust as the application language that also have this idea of not bundling chromium but instead using the system webview for rendering HTML and JavaScript.
You can find a bunch of different approaches in lists like "alternatives to electron." There's some on GitHub.
I took a slightly different approach where instead of using the system web view which I thought you know is going to be inconsistent across systems and it's not going to support the latest HTML JavaScript and security features I used the assumption that the user already has chrome installed which works in a high number of cases or can download and install it if that's not the case. predictably I suppose some people express to satisfaction that it was not using Firefox. using Firefox becomes more possible and more likely I suppose as firefox's support for the dev tools protocol achieves parity with chrome support for that.
> I kind of wonder why it took this long for someone to try this approach. It just makes a whole lot more sense on the surface.
Which is why many many people tried this approach long before Electron came along. Electron has been successful largely because it deviated from this: it swallowed the bitter pill of bundling the heavy duplicated webview in order to ensure a consistent development target.
>> I kind of wonder why it took this long for someone to try this approach. It just makes a whole lot more sense on the surface.
I believe there are dozens of projects using Webview and market themselves as a lightweight "Electron". The development for these apps takes longer (as you need to test for different webview versions/vendors) and the end product is worse as some features are shimmed or you just choose to not support them due the effort required. Of course this depends by project.
I've developed a media player and considered webview. Soon enough I've found the limitations of safari, ie webview and even firefox webview on audio/video codecs is a blocker. Next was indexeddb and some random performance issues. It didn't took long to realise that it's hard to compete with native apps using Electron but its even harder using different webview versions.
I will consider webView again when WASM gets native DOM access and audio/video decoding becomes available (performance wise) in WASM.
The dirty little secret is that a software project with two different native codebases is often cheaper, with a faster time-to-market, than a cross platform browser-tech based solution that had just one codebase.
The BS of the browser tech world compounded over time and a million edge cases eventually eats up all the one-codebase, cross-platform savings.
What we need is a subset of HTML features that are known to work across all major browser. From Firefox, Safari /WebKit on other MacOS, and Chrome.
Or something that compiles to the compatible HTML code for those platform with 100% accuracy.
Unless that happens, you are dealing with godzillion number of edge cases in hundreds of different System Browser.
But even if such language or subset of feature exist, it will still not be as good or as be limited compared what is possible on top of current Electron. Which means from a user perspective, the Dev is optimising for something they rarely cares about; download size. Given memory CPU and Memory usage to electron.
The problem is not all current Web Standards are supported across all System Browsers. So the Subset is meant as lowest common factor. How do get anything to work when you have customers using IE?
It didn’t take long, this approach was around years before Electron came in. But the reason Electron picked up is precisely the stable APIs with cross-platform compatibility that comes from bundling chromium. You won’t get that here.
It will, but browser diversity is not what you want for an electron-style app. The whole point of Electron is that it's a browser you control--you set the site permissions, you control the version and updates, you know exactly what it supports. That's what makes longer-lasting, more native experiences possible with a low development cost.
I think my best-case scenario would be an Electron replacement that is not so bloated and also keeps parts of Firefox well-funded and people looking at compatibility.
Elevating a new stack is a good consolation prize.
It seems every couple of months there is an alternative to Electron posted here which uses system webviews. A quick search turns up a wide variety of them.
Well Windows is basically Chrome so your diversity will basically be Chrome & WebKit which is what you have anyway. And having all the fun of web development against multiple browsers with all the headaches of desktop distribution is kind of the worst of all worlds. It'll be interesting to see if this approach finds success (since it may cut down on the size of your binary) but it could easily get abandoned if any significant projects start using this as frontend devs create pressure to simplify the stack across platforms.
On Linux it's going to pull in GTK which means a not great experience on KDE (even though KDE does have WebKit hooks like GTK does).
> so your diversity will basically be Chrome & WebKit which is what you have anyway
Right now many web apps can brush off doing QA in Safari (and Firefox for that matter). Most regular websites don't have issues, but most of those don't need any vendor-specific fixes in the first place. I've used multiple complex web apps that just break unceremoniously unless you're using Chrome.
If web apps that are complex enough to warrant a desktop app are forced to support Safari, that lifts awareness for the entire web/JS ecosystem. That's a win, if it plays out that way.
> having all the fun of web development against multiple browsers
The difficulty here has gone sharply down since all major browsers are evergreen now (and IE is all-but-gone). Not to mention the availability of tooling like Babel.
> Right now many web apps can brush off doing QA in Safari (and Firefox for that matter). Most regular websites don't have issues, but most of those don't need any vendor-specific fixes in the first place. I've used multiple complex web apps that just break unceremoniously unless you're using Chrome.
There's billions of active iPhones out there & they ALL use WebKit for webviews as no other browser engine is allowed for iOS apps. MacOS desktop apps are a fraction of that number of users & web view apps are a fraction of that. I just I don't follow this reasoning unless you're saying that this might gain traction across developers who might go & fix the bugs. That might help WebKit (although questionable). It's never going to help Firefox which is the sole remaining engine that doesn't have a strategic play to try to encourage compatibility.
> The difficulty here has gone sharply down since all major browsers are evergreen now (and IE is all-but-gone). Not to mention the availability of tooling like Babel.
If that were actually true, would you be concerning yourself with getting devs to do more QA on Safari or Firefox?
The truth is that it's still hard to write web apps cross-browser and will likely remain hard indefinitely as long as there are > 1 web engines. The point of the web was to make you OS agnostic (ironically recreating the same pattern in desktop development of 2 or 3 major engines as with 2 or 3 major OSes) but with this tech you're combining both dealing with the browser-specific issues & OS-specific issues. Additionally, you're now not only having to maintain compatibility across 3 engines, but all the various versions popular versions of the OS distribution will have shipped. I really struggle to see the value of this approach if I'm concerned with shipping the app & lowering my development costs & increasing velocity to shipping useful features.
As far as I know that's not accurate and the default webview on Windows is still the old Edge based one, not Chromium.
The plans here are somewhat unknown I think, but right now you have to force the user to download and install (or bundle an installer) for WebView2 or Edge beta channel if you don't want to support old Edge.
Gtk-webkit is also quite different from Safari, you can't assume they are just the same.
Starting from 20h2 version Edge is based on Chromium. And I remember when that rolled out plenty of people were irked by Microsoft simply automatically installing it (the new Edge Canary) without asking for permission.
I kind of wonder why it took this long for someone to try this approach. It just makes a whole lot more sense on the surface.
Interesting thing I just thought about: since users don't get to choose their system webview, I wonder if this has the potential to broaden browser diversity a little bit?