Cutting the fat

CUTTING THE FAT When to use Ajax and when to
reload

@theophani TIFFANY CONROY All content licensed under http://creativecommons.org/licenses/by-nc-sa/3.0/ My name
is Tiffany, and I work as an interaction designer and front-end developer at SoundCloud.

CUTTING THE FAT When to use Ajax and when to
reload When to use Ajax and when to reload ======================= This is a talk about being pragmatic. This is a talk is about how to reduce the technical complexity of web applications. This talk is about using interaction design to help make the choice between using Ajax and reloading the page.

Last summer there was a conference called “Throne of JS”
The conference had a very speciﬁc premise:

“Throne of JS … is … focused on answering the
question on every web developer’s mind: “Throne of JS is […] focused on answering the question on every web developer's mind: which framework?” Yes: rich, interactive web applications are now commonplace. There are so many frameworks for building these rich client applications, and the debate among developers is which of these frameworks to use.

“Throne of JS … is … focused on answering the
question on every web developer’s mind: which framework?” “Throne of JS is […] focused on answering the question on every web developer's mind: which framework?” Yes: rich, interactive web applications are now commonplace. There are so many frameworks for building these rich client applications, and the debate among developers is which of these frameworks to use.

Question the premise. The premise of the question “which framework”
is based on the idea that in a modern web application, the page should never be reloaded. But as designers and developers, we need to step back, and ask ourselves: why don’t we want to reload the page? When *can* we reload the page, and What is the beneﬁt of Ajax? What is the cost of using Ajax? Frameworks help up to solve technical problems, but can we ever avoid those problems entirely?

What to expect First I’m going to deﬁne some terminology.
Then we examine the beneﬁts and problems of fancy single-page web applications. Lastly, I want to show you one way of approaching your applications as a designer and developer that can help you reduce complexity, even before you choose a framework. Let’s get started.

What to expect • Terminology First I’m going to deﬁne
some terminology. Then we examine the beneﬁts and problems of fancy single-page web applications. Lastly, I want to show you one way of approaching your applications as a designer and developer that can help you reduce complexity, even before you choose a framework. Let’s get started.

What to expect • Terminology • Benefits and costs of
Ajax First I’m going to deﬁne some terminology. Then we examine the beneﬁts and problems of fancy single-page web applications. Lastly, I want to show you one way of approaching your applications as a designer and developer that can help you reduce complexity, even before you choose a framework. Let’s get started.

What to expect • Terminology • Benefits and costs of
Ajax • A pragmatic approach First I’m going to deﬁne some terminology. Then we examine the beneﬁts and problems of fancy single-page web applications. Lastly, I want to show you one way of approaching your applications as a designer and developer that can help you reduce complexity, even before you choose a framework. Let’s get started.

AJAX vs RELOAD A quick recap for everyone on what
Ajax is and what we use it for, and how it differs from a page reload.

In the early 90s, all web pages were static, and
if you wanted to see new content, you followed a link or clicked on a form submit button …

In the early 90s, all web pages were static, and
if you wanted to see new content, you followed a link or clicked on a form submit button … … and the whole page would reload.

Whole page reloads In the early 90s, all web pages
were static, and if you wanted to see new content, you followed a link or clicked on a form submit button … … and the whole page would reload.

In the second half of the 90s …

JavaScript … after JavaScript was introduced, a JavaScript object called
…

JavaScript and XMLHttpRequest … XMLHttpRequest was introduced, which allowed content
to be sent and loaded …

Asynchronous JavaScript and XMLHttpRequest asynchronously, without reloading the whole page.

Asynchronous JavaScript and XMLHttpRequest In 2005, the technique of loading
content asynchronously with JavaScript was given the name Ajax. Around that time Google released Gmail and Google Maps. (ps to big know-it-alls reading these speaker notes: yes, I know the X originally stood for XML, but making that distinction would just complicate things.)

Just one part reloads Ajax allowed us to change the
page content without a full page reload. But the more exciting use case was to make the page send and load content based on user interaction.

Let’s look at an example using Facebook messages.

I start to type a reply.

In my reply, I enter a web address.

Without me doing anything, and without interrupting me typing, the
Facebook app uses Ajax to load information about the link I entered, and adds it to my message.

When I send the message, the app uses Ajax to
send my message without reloading the page. Ajax allowed us to maintain the context of what the user was doing, without reloading the page.

Ajax → maintains context The web stopped being a collection
of static web sites with ﬁxed content, and allowed us to exchange data with the web server without reloading the page. Now, we rely heavily on Ajax to control the user's experience. To reiterate what I said before: Ajax allows us to maintain the context of what the user was doing, without reloading the page.

CONTEXT What do I mean by “context”?

“Context” is the answer to the question: where are you,
and what are you doing. By this I am not talking about where you physically are while using a web application, but where you are *within* the web application, and what you are doing there.

User Profile News Feed Store Checkout Slideshow Some examples of
different contexts could be: user proﬁle, a news feed, the checkout of a store, a slideshow. Within each of these contexts, we may not want to reload the page. Each distinct context could be its own “single-page app”, and a plain old browser refresh could be used to transition between these contexts.

But two different instances of the same “screen” are in
fact two different contexts. For example: The timeline of two different people are two different contexts.

Two separate contexts But two different instances of the same
“screen” are in fact two different contexts. For example: The timeline of two different people are two different contexts.

✓ Multiple parts of the same workﬂow can all happen
in the same context. In this example, the context is ﬁlling out a form. While the user is in this context, we want to control the user’s experience as much as possible, so that the user doesn’t lose context.

✓ Same context Multiple parts of the same workﬂow can
all happen in the same context. In this example, the context is ﬁlling out a form. While the user is in this context, we want to control the user’s experience as much as possible, so that the user doesn’t lose context.

Your application may have lots of little components and features
and bits of information, and you need to understand out how they all relate to each other.

Then you need to deﬁne which components go together.

So when I say a context, I mean these different
groupings of related bits organized into meaningful groupings. Designers need to understand, deﬁne and communicate the distinct contexts in any rich web application, being clear about the where boundaries between contexts are; where one stops and the next one starts.

When we make a transition from one context to the
next, we exit the boundary of one context and enter another. Ajax could be used to control the transition between contexts, if the designer and the developer can negotiate it by balancing the beneﬁts against the technical costs.

Designers need to understand, deﬁne and communicate the distinct contexts
in any rich web application, being clear about the where boundaries between contexts are; where one stops and the next one starts.

Ajax → maintains context So, Ajax allows us to maintain
context for users.

Lots of Ajax → “Fat Client” Web applications that use
a lot of Ajax are sometimes called “fat client”

FAT vs THIN What is a “fat” client? What do
we mean by a “fat client”. How is it different from a “thin client”

Client = code in the browser SERVER ----[ network ]---->
CLIENT ## Server and client In this discussion, the code that runs in a web browser is the “client”, in contrast with code that runs on a web server.

Thin client DATA + ALL THE LOGIC -------------------[ network ]---------------->
PRESENTATION ## Thin Client [ DATA ] + [ ALL THE LOGIC ] ----[network]----> [PRESENTATION] All the logic is performed on the server-side. Static pages that rely on page reload are THIN. The browser, AKA the client, just displays the content.

Fat client DATA -------------------[ network ]----------------> ALL THE LOGIC +
PRESENTATION ## Fat Client [ DATA ] + [ ALL THE LOGIC ] ----[network]----> [PRESENTATION] All the logic is performed on the client-side. The server sends static content over the network, and all the logic is done in the browser AKA the client. These fat clients rely on a lot of JavaScript and Ajax.

“Native Experience” People also talk about a “native experience” on
the web. Ajax allows us to emulate a “native experience” on the web, meaning, it feels like an application not a web site.

“Native Experience” and controlling transitions One of the most basic
features of a ”native” experience is the controlled transitions between contexts, for example, the sliding left and right between screens. On the web, if you reload the page, you can’t deﬁne the transition. In every browser I’ve used, the transition is the same: the page goes white, the spinner spins, and then the page loads. Reloading can be jarring. We can get around this if we load the content with Ajax, and then transition to the new context. We call these sort of sites that exclusively use Ajax “single-page apps”.

SINGLE-PAGE APPS Awesome or not awesome? So called “single-page apps”
are very sexy these days. But people have begun to see that single-page apps are also nasty beasts that bring all sorts of technical complications with them.

SINGLE-PAGE APPS Trying to recreate a native experience brings with
it all kinds of technical problems. One upon a time, if the URL in the browser changed, the page reloaded. As we made rich applications using Ajax, we ran into the problem that we could reach a state in the application that we could not link to. We started to talk about how to “deep-link” to a context within an application. This wasn’t (usually) a problem before we started using Ajax.

SINGLE-PAGE APPS back button Trying to recreate a native experience
brings with it all kinds of technical problems. One upon a time, if the URL in the browser changed, the page reloaded. As we made rich applications using Ajax, we ran into the problem that we could reach a state in the application that we could not link to. We started to talk about how to “deep-link” to a context within an application. This wasn’t (usually) a problem before we started using Ajax.

SINGLE-PAGE APPS deep-linking back button Trying to recreate a native
experience brings with it all kinds of technical problems. One upon a time, if the URL in the browser changed, the page reloaded. As we made rich applications using Ajax, we ran into the problem that we could reach a state in the application that we could not link to. We started to talk about how to “deep-link” to a context within an application. This wasn’t (usually) a problem before we started using Ajax.

SINGLE-PAGE APPS deep-linking #! hash-bang back button Trying to recreate
a native experience brings with it all kinds of technical problems. One upon a time, if the URL in the browser changed, the page reloaded. As we made rich applications using Ajax, we ran into the problem that we could reach a state in the application that we could not link to. We started to talk about how to “deep-link” to a context within an application. This wasn’t (usually) a problem before we started using Ajax.

SINGLE-PAGE APPS deep-linking JavaScript router #! hash-bang back button Trying
to recreate a native experience brings with it all kinds of technical problems. One upon a time, if the URL in the browser changed, the page reloaded. As we made rich applications using Ajax, we ran into the problem that we could reach a state in the application that we could not link to. We started to talk about how to “deep-link” to a context within an application. This wasn’t (usually) a problem before we started using Ajax.

SINGLE-PAGE APPS deep-linking JavaScript router #! hash-bang back button window.history
Trying to recreate a native experience brings with it all kinds of technical problems. One upon a time, if the URL in the browser changed, the page reloaded. As we made rich applications using Ajax, we ran into the problem that we could reach a state in the application that we could not link to. We started to talk about how to “deep-link” to a context within an application. This wasn’t (usually) a problem before we started using Ajax.

SINGLE-PAGE APPS deep-linking JavaScript router #! hash-bang back button frameworks
window.history Trying to recreate a native experience brings with it all kinds of technical problems. One upon a time, if the URL in the browser changed, the page reloaded. As we made rich applications using Ajax, we ran into the problem that we could reach a state in the application that we could not link to. We started to talk about how to “deep-link” to a context within an application. This wasn’t (usually) a problem before we started using Ajax.

Let’s pause for a moment: Why do we want to
use Ajax? What problems are we trying to solve with Ajax? i.e. what **beneﬁts** do we gain from using it?

Problems a fat client solves: Problems a fat client solves:
* The user can continue to interact (including simply viewing) while data is exchanged. * The designer can customize and control the loading and transition experience between contexts. * Information can be presented in a partial, incremental, incomplete form, thus context can be established while further information is retrieved. (Like a pan in a ﬁlm, setting can be established in advance of details.) * The traffic between server and client can be reduced to data separate from the presentation, reducing network use and speeding up information transmission. Most of these are design problems. Even improved speed would result in a better experience for people.

Problems a fat client solves: • Continuous interaction Problems a
fat client solves: * The user can continue to interact (including simply viewing) while data is exchanged. * The designer can customize and control the loading and transition experience between contexts. * Information can be presented in a partial, incremental, incomplete form, thus context can be established while further information is retrieved. (Like a pan in a ﬁlm, setting can be established in advance of details.) * The traffic between server and client can be reduced to data separate from the presentation, reducing network use and speeding up information transmission. Most of these are design problems. Even improved speed would result in a better experience for people.

Problems a fat client solves: • Continuous interaction • Transitions
between contexts Problems a fat client solves: * The user can continue to interact (including simply viewing) while data is exchanged. * The designer can customize and control the loading and transition experience between contexts. * Information can be presented in a partial, incremental, incomplete form, thus context can be established while further information is retrieved. (Like a pan in a ﬁlm, setting can be established in advance of details.) * The traffic between server and client can be reduced to data separate from the presentation, reducing network use and speeding up information transmission. Most of these are design problems. Even improved speed would result in a better experience for people.

between contexts • Incremental loading Problems a fat client solves: * The user can continue to interact (including simply viewing) while data is exchanged. * The designer can customize and control the loading and transition experience between contexts. * Information can be presented in a partial, incremental, incomplete form, thus context can be established while further information is retrieved. (Like a pan in a ﬁlm, setting can be established in advance of details.) * The traffic between server and client can be reduced to data separate from the presentation, reducing network use and speeding up information transmission. Most of these are design problems. Even improved speed would result in a better experience for people.

between contexts • Incremental loading • Reduced network traffic Problems a fat client solves: * The user can continue to interact (including simply viewing) while data is exchanged. * The designer can customize and control the loading and transition experience between contexts. * Information can be presented in a partial, incremental, incomplete form, thus context can be established while further information is retrieved. (Like a pan in a ﬁlm, setting can be established in advance of details.) * The traffic between server and client can be reduced to data separate from the presentation, reducing network use and speeding up information transmission. Most of these are design problems. Even improved speed would result in a better experience for people.

between contexts • Incremental loading • Reduced network traffic → DESIGN PROBLEMS (mostly) Problems a fat client solves: * The user can continue to interact (including simply viewing) while data is exchanged. * The designer can customize and control the loading and transition experience between contexts. * Information can be presented in a partial, incremental, incomplete form, thus context can be established while further information is retrieved. (Like a pan in a ﬁlm, setting can be established in advance of details.) * The traffic between server and client can be reduced to data separate from the presentation, reducing network use and speeding up information transmission. Most of these are design problems. Even improved speed would result in a better experience for people.

Problems a fat client causes: Problems a fat client causes:
* Routing becomes the responsibility of the client-side script instead of the browser --> deep-linking * History management becomes the responsibility of the client-side script instead of the browser --> back button * Caching (what and for how long) becomes the responsibility of the client-side script instead of the browser --> reusing previously-fetched data and templates to achieve improved speed * Improvements from the speedy transmission are eroded by computation within the client i.e. instead of doing computation in an environment we can control (the server) we push it to the client, which is notoriously out of our control. These are technical problems. From a design perspective, these become trade-offs that are “someone else’s problem”.

Problems a fat client causes: • Routing Problems a fat
client causes: * Routing becomes the responsibility of the client-side script instead of the browser --> deep-linking * History management becomes the responsibility of the client-side script instead of the browser --> back button * Caching (what and for how long) becomes the responsibility of the client-side script instead of the browser --> reusing previously-fetched data and templates to achieve improved speed * Improvements from the speedy transmission are eroded by computation within the client i.e. instead of doing computation in an environment we can control (the server) we push it to the client, which is notoriously out of our control. These are technical problems. From a design perspective, these become trade-offs that are “someone else’s problem”.

Problems a fat client causes: • Routing • History Problems
a fat client causes: * Routing becomes the responsibility of the client-side script instead of the browser --> deep-linking * History management becomes the responsibility of the client-side script instead of the browser --> back button * Caching (what and for how long) becomes the responsibility of the client-side script instead of the browser --> reusing previously-fetched data and templates to achieve improved speed * Improvements from the speedy transmission are eroded by computation within the client i.e. instead of doing computation in an environment we can control (the server) we push it to the client, which is notoriously out of our control. These are technical problems. From a design perspective, these become trade-offs that are “someone else’s problem”.

Problems a fat client causes: • Routing • History •
Caching and garbage collection Problems a fat client causes: * Routing becomes the responsibility of the client-side script instead of the browser --> deep-linking * History management becomes the responsibility of the client-side script instead of the browser --> back button * Caching (what and for how long) becomes the responsibility of the client-side script instead of the browser --> reusing previously-fetched data and templates to achieve improved speed * Improvements from the speedy transmission are eroded by computation within the client i.e. instead of doing computation in an environment we can control (the server) we push it to the client, which is notoriously out of our control. These are technical problems. From a design perspective, these become trade-offs that are “someone else’s problem”.

Caching and garbage collection • Computation is outside of our control Problems a fat client causes: * Routing becomes the responsibility of the client-side script instead of the browser --> deep-linking * History management becomes the responsibility of the client-side script instead of the browser --> back button * Caching (what and for how long) becomes the responsibility of the client-side script instead of the browser --> reusing previously-fetched data and templates to achieve improved speed * Improvements from the speedy transmission are eroded by computation within the client i.e. instead of doing computation in an environment we can control (the server) we push it to the client, which is notoriously out of our control. These are technical problems. From a design perspective, these become trade-offs that are “someone else’s problem”.

Caching and garbage collection • Computation is outside of our control → TECHNICAL PROBLEMS Problems a fat client causes: * Routing becomes the responsibility of the client-side script instead of the browser --> deep-linking * History management becomes the responsibility of the client-side script instead of the browser --> back button * Caching (what and for how long) becomes the responsibility of the client-side script instead of the browser --> reusing previously-fetched data and templates to achieve improved speed * Improvements from the speedy transmission are eroded by computation within the client i.e. instead of doing computation in an environment we can control (the server) we push it to the client, which is notoriously out of our control. These are technical problems. From a design perspective, these become trade-offs that are “someone else’s problem”.

Design problems a fat client solves: Let’s look again at
the interaction design problems that a fat client can solve.

Design problems a fat client solves: • Continuous interaction Let’s
look again at the interaction design problems that a fat client can solve.

Design problems a fat client solves: • Continuous interaction •
Transitions between contexts Let’s look again at the interaction design problems that a fat client can solve.

Transitions between contexts • Incremental loading Let’s look again at the interaction design problems that a fat client can solve.

Transitions between contexts • Incremental loading Maintain context Let’s look again at the interaction design problems that a fat client can solve.

Transitions between contexts • Incremental loading Maintain context Establish context Let’s look again at the interaction design problems that a fat client can solve.

Ajax → controls context Ajax allows us to establish and
maintain the context for the user, without jarring reloads. BUT …

Ajax → adds complexity Ajax also introduces all lot of
technical complications. This is complication that you as designer don’t have to think about, but creates a lot of extra work for developers. And that extra work might not be worth it. How can designers help developers be pragmatic about when to use Ajax and when not to?

Transitions between contexts • Incremental loading When we look at the problems that a fat client can solve, we can ask which one is the most important? -> Continuous interaction is the primary reason we use Ajax. We use it during form validation, live chatting, commenting while watching a video

Transitions between contexts • Incremental loading Most important When we look at the problems that a fat client can solve, we can ask which one is the most important? -> Continuous interaction is the primary reason we use Ajax. We use it during form validation, live chatting, commenting while watching a video

Continuous interaction Continuous interaction is the primary reason we use
Ajax. We use it during form validation, live chatting, commenting while watching a video, performing a search.

Transitions between contexts • Incremental loading So what beneﬁt of Ajax is the least important? -> Transitions between contexts

Transitions between contexts • Incremental loading Least important So what beneﬁt of Ajax is the least important? -> Transitions between contexts

New context? → Reload! When the user switches to a
new context, the page can just be reloaded.

NO TRANSITIONS!? What about transitions? Transition are nice to have
but not essential for most cases.

“The cool kids are doing it!” You might say: BUT
BUT BUT … all the cool kids are doing it! My product is also cool! Maybe, but some of the cool kids still use page reload.

Case study: Github

On github.com: a proﬁle is a context, the repo ﬁle-browser
is a context, and each pull request is one context, with page reloads in between.

As we saw, Github is a collection of single-page apps,
one per context, with page reloads in between.

AJAX vs RELOAD? So, when should you use Ajax and
when should you use reload? Let’s recap:

Limit your demand for Ajax Limit your demand for asynchronous
communication to only within the contexts that matter. Only use it when you can afford it. “Single-page apps” are often made as single-page apps for the wrong reasons. I have made this mistake. Where is our need for Ajax? -- at the interaction points.

Find the interaction points Find the interaction points. Everywhere the
user will click, type, hover or otherwise engage with the interface. For each, ask: what beneﬁt vs cost does Ajax have at that interaction point? Can we afford the cost? Where can we eliminate our need for Ajax?

Find the context boundaries Know where the boundaries between contexts
are, and communicate them to developers, because these are important for designing the application architecture. Using Ajax in at these interaction points give us the least beneﬁt. Lastly, for developers:

Does your framework minimize your technical costs? Does your framework
help solve the design problems while minimizing the technical costs?

Single-page apps? YES: we can help our users maintain context
so they can focus without jarring interruptions. These means: use Ajax for each context, and each of those is its own single-page app. Monolithic apps? NO: or at least, they aren’t nearly as useful as they are technically hard, and you should be very clear of the costs. You may get so much beneﬁt, or have so many people working in the problems, that you can overcome the costs. Your *entire web app* does *not* need to be one single front-end app.

Single-page apps? → YES Single-page apps? YES: we can help
our users maintain context so they can focus without jarring interruptions. These means: use Ajax for each context, and each of those is its own single-page app. Monolithic apps? NO: or at least, they aren’t nearly as useful as they are technically hard, and you should be very clear of the costs. You may get so much beneﬁt, or have so many people working in the problems, that you can overcome the costs. Your *entire web app* does *not* need to be one single front-end app.

Single-page apps? → YES Monolithic apps? → NO Single-page apps?
YES: we can help our users maintain context so they can focus without jarring interruptions. These means: use Ajax for each context, and each of those is its own single-page app. Monolithic apps? NO: or at least, they aren’t nearly as useful as they are technically hard, and you should be very clear of the costs. You may get so much beneﬁt, or have so many people working in the problems, that you can overcome the costs. Your *entire web app* does *not* need to be one single front-end app.

QUESTIONS? and discussion @theophani speakerdeck.com/theophani/cutting-the-fat

THANKS! All content licensed under http://creativecommons.org/licenses/by-nc-sa/3.0/

Cutting the fat

Cutting the fat

More Decks by Tiffany Conroy

Other Decks in Design

Featured

Transcript