Dion Almaer

Software, Development, Products

Web Platform

The holy grail of a Web SDK

As I sit here watching WWDC, I get that same feeling. Everything is vertically integrated. The layering of the SDKs, with much that is shared, and then differences for iOS vs. Mac vs. tv, etc. You can’t hope but look at the Web and think “oh man, the paradox of choice over here… with so much evolving and changing to keep up with!”

The old analogy comparing dictatorships vs. democracies continues to ring somewhat true.

I have made the mistake in the past of wanting to chase the holy grail of a Web SDK that has the benefits of a unified stack that takes care of everything. However, I have learned that this is a holy grail in the sense that it can’t actually exist on the Web in the form that you think of.

The quest is wrong. We shouldn’t try to be like Apple, and compete in the same way with a history and ecosystem that isn’t setup to succeed in that form.

Well, what is the right quest?

  • We should embrace the ecosystem, and the players lower in the stack (e.g. browsers shipping the web platform) should embrace the fact that so much evolution happens on top. Plan for this. Be malleable and modular, and allow for shifts and changes. Most of all, be enablers for the stack on top.
  • The web platform can absolutely do better with undergrounding gains that come from the evolution in user space. The Web is a meta-platform that ships with browsers. Anything in user space is fairly costly, but has the trade-off where it can evolve quickly (downloading libraries / updating libraries).
  • It is a fair criticism that the Web has many primitives baked in for documents, but leaves much of app development to user space. If you build something truly app-y, you can end up somewhat frustrated. It’s improving all the time, and there are things that are amazing about building web apps, but there is a ton to improve here.
  • In general, there is value in seeing what primitives are needed to help the various types of experiences that users choose on the Web. It’s so broad, and the needs for content, commerce, apps, and games are different and yet share a lot.

I am seeing a lot of progress in areas that push us forward:

  • WebAssembly as a lot level runtime for client AND server that can change the game, opening up a TON of room for the ecosystem to innovate.
  • Massive CSS improvements with container queries and new helpful syntax bumps, and losing the need for preprocessors for obvious changes (this is all good undergrounding!!)
  • Apple hiring and catching up with app related APIs (e.g. Web Push, making PWA not be somewhat broken). There is still so much that is needed here, both in WebKit catching up and also the entire Web pushing forward. Dogfooding with Active Recall is very humbling!

Let’s lean in. Platform folks work together and build hooks and capabilities that those on top can innovate with. App devs and framework/library folks, share what you need. Together, push on where we want the Web to go and use the fact that it isn’t run by one company.

App Mode

What would <html app> do? 🤔

The Web can continue to evolve by taking the long term view and chipping away with adding necessary functionality, and eliding some of the mistakes (albeit hard).

However, I have been pondering if we should be exploring some bigger bets:

  • First, looking at certain verticals (content, commerce, apps) and prioritizing web platform features and tooling to make them successful holistically
  • Second, breaking out of the “browser” and be progressive all the way to the native platforms by adding escape hatches

Now, let’s take another look at building the UI for apps on the Web in a very webby way, embracing what’s good about HTML, CSS, and JavaScript, but really re-thinking it for apps.

Don’t let evolution get in the way

How connected are we? Very.

Evolution is resilient and special. Pair it with immense time frames, and life will find a way, even as many experiments will be left on the way side. One of the strengths of the Web is how it evolves as a platform, but it can be a double edge sword.

We can’t expect to constantly evolve while keeping the entire history front and center. We have fragments in our DNA that go back to the birth of life, but much of it is very much hidden and ignored.

We can embrace evolution that allows us not to keep the baggage around. If you look at a hello world web app, that supports a decent set of browsers, you will find yourself setting up resets of sorts, meta tags, and a series of cruft. This cruft creates cognitive overhead for developers.

In hindsight, we will always make mistakes in the platform and come up with solutions. One simple example that we all know, is that of box-sizing.

I don’t want the default content-box model, and instead agree with Paul from 2012, and many since, who recommend we go all in and flip the default to: box-sizing: border-box;

In 2021, the vast majority of your users will be coming at you via an evergreen modern browser. This is huge, and allows us the ability to flip to a modern-first approach of development that doesn’t tax all of our users for the lowest common denominator. Finally, we can speed up evolution!

While it’s phenomenal that old content sitting on the Web can still be browsed (but of course, this isn’t always true, plenty of old content breaks), it also means that browsers are constrained. They can’t just flip the defaults because *new* content is being developed using new ways, thus we are a lil stuck.

How could browsers do the best job they can at rendering all content, including great content written many moons ago, whilst allowing for new content?

Multiple Rendering Paths

Remember quirksmode? I still see developers accidentally finding themselves confused, because they have put their document into quirks vs. standard mode.

We also have AMP, which forces a constrained mode, optimized for documents keyed off of:

<html ⚡ lang="en">

So, what if we took a leaf out of these books, and had something like:

<html app lang="en">

This could give us a way to tell the browser that we want a new world, one optimized for apps rather than documents. A way to trigger a reset. This would give us a chance to re-think building web apps that still uses HTML, CSS, and JavaScript and feels very webby, but in a sane way for the app use case that is quite different to content.

What could we see?

App-oriented Layout

I love our Grid and Flexbox world, but we could also look into constrained layout modes, working closely with tooling vendors as the Android team did when they shipped their most recent layout systems.

App-oriented Components

Take another look through the HTML spec and the various elements you have built in. I bet you will find some elements that you have never used, such as <aside>. There are a ton of document focused elements, which is great!

But we stopped. While it’s great to abstract and create our own components, either through Web Components themselves, or via framework abstractions that spew out div soup, why haven’t we undergrounded the great learnings from the ecosystem? My kingdom for tabbed UI, and the good ole UITableView! Fortunately, I am seeing more exploration and appetite for this!

App-oriented Data Flow

When you are building rich web apps, and you try to really build your app offline-first, with optimistic UI, you run into a lot of issues around data and syncing.

We just saw James Long dig into the mess that is IndexedDB and the lack of a portable WebSQL/SQLite. What can we do to really enable experiences that can rely on data?

Are we forcing it? What about WebAssembly?

But wait a minute, are we barking up the wrong tree here? Maybe the solution is to keep HTML, the DOM, and friends for the content Web? Or maybe content++… allowing for interactivity but not as far as full on apps!

Flutter using the Web as a target

With WebAssembly, we could keep URLs, but have them toss back apps that are WebGL renderers! We could just use Flutter for Web as an example. This would allow many runtimes to evolve on top, and bobs your uncle.

While this may make sense for some use cases, and I love that Flutter developers have a path to Web, I don’t think that we should cede all apps to this model, as it naturally ends up with blobs at the end of URLs, and we lose a lot of what is great about the HTML/CSS/JS web.

I know that it is quite niche, but I love how Chrome Extensions can compose and hack the Web. This is a super power akin to an after market for car parts. Someone can create functionality that helps you with another services, and I would like more of that, not less. The Web is fantastic at streaming UI, and we can do more to make this all work well with apps, versus the years of fighting against it and downloading the world before rendering anything.

Thus, I find myself dreaming of layers, or adding the primitives that allow for evolution, but also having another go at baking app UI right into the web as a first class citizen, instead of it being a render target with so many foot guns and uncanny valleys. I dream of a studio of tools, aligned with the platform, to make this possible working hand in hand.

And then I wake up to see debates around a simple alert(), and how hard it is to make changes, and I once again ponder escape hatches that allow us to make bold changes.

It’s complex and hard to make some moves, as when you have a platform you need to make sure that both:

  • developers have a happy path, know when they are doing something sub-optimal and how to fix it, and make that happy path the one that doesn’t lead you there
  • and, we minimize the attack surface to make it as hard as possible for bad actors to do harmful things to users and businesses.

And there’s often the rub, as the same capabilities and loopholes can be used for both.

Apps were born on a document platform, and we are still lacking a holistic web-first app platform that could unlock experiences that have great attributes from the web, but are also more fun and productive for developers to build.

What is in your app mode?

Time to Embrace Escape Hatches on the Web Platform?

Progressive Web Apps aren’t reaching their potential at scale. I just spoke about how I think it’s important for the Web to push on its own vision as a platform, but as a universal app platform (one of the key segments, along with content and commerce) it isn’t hitting the mark, mainly on mobile. What if we rethink the web platform, and embrace the “progressive” quality to allow for escape hatches on top of web standards as an interop layer for maximal reach?

The default mode of evolution on the web is deliberate. This gives us a nice eventual consistency across multiple implementations, but  there are times in which it is held back too much and at those times I think it is important for us to break out of the mold, branch off and prove value, while always looking to merge back in the future. We are at one of those times, and here is why our current situation isn’t working for users and developers (and thus the platform).

User Expectations

People go to app stores to find apps. On desktop, the browser as a virtual machine kinda works, and there is enough real estate for the browser chrome to be a feature not just a bug. On mobile that isn’t the case. And expecting users to understand Add to Home Screen is swimming upstream. And of course, there are plenty of uncanny valley situations when using a PWA. It doesn’t come across when you migrate devices. It doesn’t show up across an ecosystem (e.g. widgets, watches, notifications on some platforms, etc).

Developer Expectations

As a developer, knowing that I can’t hit the user bar already has me thinking about how far I can go. I sure as hell don’t want to rewrite a mobile app to target multiple platforms, and LOVE the idea of not reinventing the wheel, but I need to not be locked out of key platform capabilities and user flows.

Add to that the notion of FOMO. When iOS and Android release their new SDKs, how long will it take for me to get access via a web standard feature? Can I risk being locked out of $SOME_COOL_NEW_FEATURE?

And will Apple ever implement what I need? They have shown, through action, that they have a different vision for the Web, one that is much more focused on the content Web. Given their business and app strategy, it isn’t being evil, it makes sense. They want developers building on their native stack and offering experiences that are differentiated.

We have all tried to push on this for some time, but maybe it’s time for a different tact. Instead of slowly doing all we can to change the calculus for Apple to implement something (e.g. finding ways for their internal teams to request a feature, or getting adoption of something so that it works poorly or slowly, such as YouTube supporting Web Components) let’s think different.

Web as an Interop Layer

It’s interesting to look at iMessage and SMS as another path. The SMS standard acts as an interop layer, where you are able to text between iOS and Android. It doesn’t have all of the fancy features… but it allows for fast iterations when each platform is talking to another device on it’s own system.

This is far from perfect. How frustrating is it to react to a text on iOS, but because someone on the thread is on Android, you get another text with: Liked “repeat of the message”. I see this and conspiracy theories start up in my head…. Surely you could write the Regex to see the pattern and just show the reaction on that message :/

And then there is Spotify and podcasts. Originally, you were very much locked in, but recently Spotify supported RSS and your ability to monetize your own way if you wish. If you can offer choice to developers where they can get different levels of integration based on the direction they go (doing extra work for a platform vs. open standards) that is a real trade off with real choice. Once again though, far from perfect, and I am frustrated with Spotify sucking up “exclusives”, and shoving podcasts down my throat in the app when I just use it for music.

What would this look like for the Web? The standards (and really the implementation of them… which is why compat efforts will always be so important) are the interop layer that gives you the base level functionality and allows you to reach every user with a device that implements a web client. We use progressive enhancement as always to do the right thing to make the experience work as well as it can for the given device and modalities and browser and version that a user is browsing your site with.

But we progressively enhance further, breaking out of the sandbox as needed. The starting point is on getting browser vendors onboard, but if to bogged down you aren’t stuck. We lean into Capacitor type approaches, letting you reach to native when needed, and getting you into the stores that don’t support you with a simple URL. Platforms can give you extras and you can choose to put in the extra work to enhance to them if the ROI is there. We learn from this evolution and always push to eventually bake it into the standards. Just like how I wish that reactions get baked into new versions of texting standards, and how chat apps could eventually get interop (fun fact: I used to be an IRC admin for the University of Minnesota!)

What does this look like?

Let’s say I build an image editor application. On ChromeOS the web is the app platform, so I want my users to be able to double click on the PNG file on their desktop and have it launch and load in the PWA. If this feature isn’t fully implemented on other browsers, give me a way to do it on ChromeOS first. It’s OK!

Now let’s say we want to bring this editor to mobile and are frustrated at the lack of support for *simple* features such as access to a file system. If one OS isn’t getting you what you need, we break off and have a wrapper API that bridges through the native SDK. We have polyfills where we can, and clean ways to use enhanced APIs that don’t break the web experience when we can’t. We can even embrace the fact that the app install step is a nod that the user cares about this experience.

Trust is important. The ephemeral nature of the web is key, and I’m all for leveling the field, but we also need user signals that don’t make UX of an API awful.

Let’s take the image editor example and say that we want to add rich font support and we want to be able to access your local system fonts. If you meet with the security and privacy team first they may point out that a web site knowing which fonts you have is just what finger printing folks would love to see. Instead, what if we protect users and when you sign in we ask users to choose which fonts they want to share…. Like sharing images via a picker vs granting access to your camera roll. You nod for a bit, and then you think through that UX.

Wait, how do I know which fonts I want to use in the future? In my editor I want users to be able to play with their design and quickly flip through fonts to see how they look with the content I have?

Now imagine that this image editor is from Adobe, and picture the user…. Logging into creative cloud which they pay for, and seeing this? Does this keep them safe? Or does it result in closing the web app and just switching to native?

We have this working now!

We a history of this evolution. On desktop with Electron, and a huge number of mobile apps embed WebViews, use custom bridges, and use products such as Capacitor.

And we are seeing success stories, particularly within certain verticals such as enterprise and publishing. I remember meeting with a huge media company in Latin America who switched from their native apps to using web + native where needed. It wasn’t just useful for their general ROI, all of their metrics went up and to the right:

  • Positive App Store ratings increased by 25%
  • App size was reduced by 70%
  • Crashes reduced by about 80%
  • Engagement increased.

But how? Well, context matters. They had a large web team and small native teams and this resulted in native always being far behind in features and the overall quality not being the best. You don’t get platform leverage for free. You still need native skills but a core group could watch over some of that and the feature teams could run faster.

And now when this years SDK comes out, they can bridge over to it and access that new branch of evolution.

Let’s consider how to keep up the pace and deliver on user and developer expectations by offering progressive enhancement down to the underlying platform vis escape hatches, and do it in a way that learns from the past (e.g. MS universal apps) allowing us to be constantly merging improvements into main, which are the web standards that we hold dear.

The right thing to do, is the right thing to do.