Mike Chambers

There has been a lot of excitement, interest and discussion around Apollo, especially since we released the public alpha on labs last week. One thing that has come up a couple of times, is confusion over what Apollo is as well as what value it provides. A lot of the discussion has focused on uncertainty about why would you want to move web applications outside of the browser.

A lot of times when this question gets answered, the answer focuses on specific Apollo features (file I/O API, working offline). While these are things that Apollo can do today, that are difficult if not impossible to do consistently in the browser, a feature based discussion doesn’t address the fundamental question of why would you want to move applications out of the browser.

I had been planning to write up my thoughts on this, and realized that I already had as part of the Apollo Pocket Guide for Flex Developers. Below is chapter one from that book, which explains what Apollo is, and what problems it is trying to solve. (You can download the entire book from here).

Note that the excerpt does contain a discussion of features, but one of the primary advantages of Apollo, which isn’t a specific feature, is that it allows applications to run outside of the browser. This is not a ding on browsers, or web technologies, and as I point out, the browser has some strong advantages that often outweigh its disadvantages.

Ultimately though, because browser based and Apollo based applications are built using the same technologies, it is possible to deploy to both platforms, taking advantages of the strengths of each. Because of this, Apollo applications compliment web applications. They do not replace them.

What is Apollo?

Apollo is a new cross-platform desktop runtime being developed by Adobe
that allows web developers to use web technologies to build and deploy
Rich Internet Applications and web applications to the desktop.

In order to better understand what Apollo enables, and which problems
it tries to address, it is useful to first take a quick look over at
the (relatively short) history of web applications.

A Short History of Web
Applications

Over the past couple of years, there has been an accelerating trend of
applications moving from the desktop to the web browser. This has been
driven by a number of factors, which include:

• The growth of the Internet as a communication medium
• The relative ease of deployment of web applications
• The ability to target multiple operating systems via the
  browser
• The maturity of higher-level client technologies, such the
  browser and the Flash Player runtime

Early web applications were built primarily with HTML and JavaScript,
which, for the most part, heavily relied on client/server interactions
and page refreshes. This page refresh model was consistent with the
document-based metaphor for which the browser was originally designed,
but provided a relatively poor user experience when displaying
applications.

However, with the maturation of the Flash Player runtime, and more
recently Ajax-type functionality in the browser, it became possible for
developers to begin breaking away from page-based application flows. In
short, developers began to be able to offer richer application
experiences via the browser. In a whitepaper from March 2002,
Macromedia coined the term Rich Internet Application (RIA), to describe
these new types of applications in browsers, which “blend
content, application logic and communications…to make the
Internet more usable and enjoyable.” These applications
provided richer, more desktop-like experiences, while still retaining
the core cross-platform nature of the Web:

Internet applications are all about reach. The promise of the web is
one of content and applications anywhere, regardless of the platform or
device. Rich clients must embrace and support all popular desktop
operating systems, as well as the broadest range of emerging device
platforms such as smart phones, PDAs, set-top boxes, game consoles, and
Internet appliances.

You can find the complete whitepaper and more information on RIAs href="http://download.macromedia.com/pub/flash/whitepapers/richclient.pdf">here.

The paper goes on to list some features that define RIAs:

• Provide an efficient, high performance runtime for
  executing code, content, and communications.
• Integrate content, communications, and application
  interfaces into a common environment.
• Provide powerful and extensible object models for
  interactivity.
• Enable rapid application development through components and
  re-use.
• Enable the use of web and data services provided by
  application servers.
• Embrace connected and disconnected clients.
• Enable easy deployment on multiple platforms and devices.

This movement toward providing richer, more desktop-like application
experiences in the browser (enabled by the Flash Player runtime, and
more recently by Ajax techniques) has led to an explosion of web
applications.

Today the web has firmly established itself as an application
deployment platform that offers benefits to both developers and end
users. These benefits include the ability to:

• Target multiple platforms and operating systems.
• Develop with relatively high-level programming and layout
  languages.
• Allow end users to access their applications and data from
  virtually any Internet-connected computer.

The growth of web applications can be seen in both the Web 2.0
movement, which consists almost entirely of web based applications and
APIs, as well as the adoption of web applications as a core business
model of major companies and organizations.

Problems with Delivering
Applications via the Browser

As web applications have become more complex, they have begun to push
the boundaries of both the capabilities of the browser and the
usability of the application. As their popularity grows, these issues
become more apparent and important and highlight that there are still a
number of significant issues for both developers and end users when
deploying applications via the browser.

The web browser was original designed to deliver and display HTML-based
documents. Indeed, the basic design of the browser has not
significantly shifted from this purpose. This fundamental conflict
between document-and application-focused functionality creates a number
of problems when deploying applications via the browser.

Conflicting UI
 
Applications deployed via the browser have their own user interface,
which often conflicts with the user interface of the browser. This
application within an application model often results in user
interfaces that conflict with and contradict each other. This can lead
to user confusion in the best cases, and application failure in the
worst cases. The classic example of this is the browser’s
Back button. The Back button makes sense when browsing documents, but
it does not always make sense in the context of an application.
Although there are a number of solutions that attempt to solve this
problem, they are applied to applications inconsistently; users may not
know whether a specific application supports the Back button, or
whether it will force their application to unload, causing it to lose
its state and data.

Distance from the Desktop

Due in part to the web security model (which restricts access to the
users machine), applications that run in the browser often do not
support the type of user interactions with the operating system that
users expect from applications. For example, you cannot drag a file
into a browser-based application and have the application act on that
file. Nor can the web application interact with other applications on
the user’s computer.

RIAs have tried to improve on this by making richer, more desktop-like
interfaces possible in the browser, but they have not been able to
overcome the fundamental limitations and separation of the browser from
the desktop.

Primarily Online
Experience

Because web applications are delivered from a server and do not reside
on the users machine, web applications are a primarily online
experience. While there are attempts underway to make offline web-based
applications possible, they do not provide a consistent development
model, they fail to work across different browsers, and they often
require the user to interact with and manage their application and
browser in complex and unexpected ways.

Lowest Common Denominator

Finally, as applications become richer and more complex and begin to
push the boundaries of JavaScript and DHTML, developers are
increasingly faced with differences in browser functionality and APIs.
While these issues can often be overcome with browser-specific code, it
leads to code that is more difficult to maintain and scale, and takes
time away from function-driven development.

While JavaScript frameworks are a popular way to help address these
issues, they can offer only the functionality provided by the browser,
and often resort to the lowest common denominator of features between
browsers to ease the development model. While this issue
doesn’t affect Flash-based RIAs, the end result for
JavaScript-or DHTML-based applications is a lowest common denominator
user experience and interaction model, as well as increased
development, testing, and deployment costs for the developer.

The fact that web applications have flourished despite these drawbacks
is a testament to the attractiveness of having a platform with a good
development model that has the ability to deliver applications to
multiple operating systems. A platform that offered the reach and
development model of the browser, while providing the functionality and
richness of a desktop application, would provide the best of both
worlds. This is what Apollo aims to do.

Introducing the Apollo
Runtime

So, what is Apollo, and how can it make web application development and
deployment better?

Apollo is the code name for a new cross-operating system runtime being
developed by Adobe that allows web developers to leverage their
existing web development skills (such as Flash, Flex, HTML, JavaScript,
and PDF) to build and deploy Rich Internet Applications and content to
the desktop.
In essence, it provides a platform in between the desktop and the
browser, which combines the reach and ease of development of the web
model with the functionality and richness of the desktop model.

It is important to step back for a second and point out what Apollo is
not. Apollo is not a general desktop runtime meant to compete with
lower-level application runtimes. This means that you probably
wouldn’t want to build Photoshop on top of Apollo.
Apollo’s primary use case is enabling Rich Internet and web
applications to be deployed to the desktop. This is a very important
but subtle distinction, as enabling RIAs on the desktop is the primary
use case driving the Apollo 1.0 feature set.

Apollo is also not a browser. While it does allow developers to use web
technologies such as HTML, and Flash to build applications, it does not
provide any default UI to the user for browsing web pages. It is
possible to build a browser on top of Apollo, but at a basic level,
Apollo is a runtime that enables developers to build end user
applications, and is not an end user application itself.

At its core, Apollo is built on top of web technologies, which allow
web developers to develop for and deploy to the desktop using the same
technologies and development models that they use today when deploying
applications on the Web.

Primary Apollo
Technologies

There are three primary technologies included within Apollo, which fall
into two distinct categories: application technologies and document
technologies.

Primary Application
Technologies

Application technologies are technologies that can be used as the basis
of an application within Apollo. Apollo contains two primary
application technologies, Flash and HTML, both of which can be used on
their own to build and deploy Apollo applications.

Flash

One of the core technologies Apollo is built on is the Flash Player.
Specifically, Apollo is built on top of Flash Player 9, which includes
the ECMAScript-based ActionScript 3 as well as the open source Tamarin
virtual machine (which will be used to interpret JavaScript in future
versions of Firefox).

You can find more information on the open source Tamarin project at on
the Mozilla website site at http://www.mozilla.org/projects/tamarin/.

Not only are all of the existing Flash Player APIs available within
Apollo, but some of those APIs have also been expanded and/or enhanced.
Some of the functionality that the Flash Player provides to Apollo
includes:

• Just-in-time Interpreted ActionScript engine for speedy
  application performance
• Full networking stack, including HTTP and RTMP, as well as
  Binary and XML sockets
• Complete vector-based rendering engine and drawing APIs
• Extensive multimedia support including bitmaps, vectors,
  audio, and video

Of course, the Flex 2 framework is built on top of ActionScript 3,
which means that you can also take advantage of all of the features and
functionality that Flex offers in order to build Apollo applications.

HTML

The second application technology within Apollo is HTML. This is a full
HTML ren•dering engine, which includes support for:

• HTML
• JavaScript
• CSS
• XHTML
• Document Object Model (DOM)

Yes, you read that right. You don’t have to use Flash to
build Apollo applications. You can build a full-featured application
using just HTML and JavaScript. This usually surprises some developers
who expect Apollo to focus only on Flash. However, at its core, Apollo
is a runtime targeted at web developers using web
technologies— and what is more of a web technology than HTML
and JavaScript?

The HTML engine used within Apollo is the href="http://www.webkit.org">open source WebKit engine.
This is the engine behind a number of browsers, including KHTML on KDE
and Safari on Mac OS X.

Why WebKit?

Adobe spent a considerable amount of time researching which HTML engine
to use within Apollo and used a number of criteria that ultimately led
them to settle on WebKit.

Open project.
Adobe knew from the very beginning that it did not want to create and
maintain its own HTML rendering engine. Not only would this be an
immense amount of work, but it would also make it difficult for
developers, who would then have to become familiar with all of the
quirks of yet another HTML engine.

WebKit provides Apollo with a full-featured HTML engine that is under
continuous development by a robust development community that includes
individual developers as well as large companies such as Nokia and
Apple. This allows Adobe to focus on bug fixes and features, and also
means that Adobe can actively contribute back to WebKit, while also
taking advantage of the contributions made by other members of the
WebKit project.

Proven technology that
web developers know. As discussed earlier, one of the
biggest problems with complex web development is ensuring that content
works consis•tently across browsers. While something may work
perfectly in Firefox on the Mac, it may completely fail in Internet
Explorer on Windows. Because of this, testing and debugging
browser-based content can be a nightmare for developers.

Adobe wanted to ensure that developers were already familiar with the
HTML engine used within Apollo, and that they did not have to learn new
all of the quirks and bugs of a new engine. Since Safari (which is
built on top of WebKit) is the default browser for Mac OS X, developers
should be familiar with developing for it.

Minimum effect on Apollo
runtime size. The target size for Apollo is between 5 and
9 MB. The WebKit code base was well-written and organized and had a
minimal impact on the final Apollo runtime size. Indeed, the current
runtime size with both Flash and HTML is just a little over 5 MB.

Proven ability to run on
mobile devices. While the first release of Apollo runs
only on personal computers, the long-term vision is to extend the
Apollo runtime from the desktop to cell phones and other devices.
WebKit has a proven ability to run on such devices and has been ported
to cell phones by both Nokia and Apple.

Primary Document
Technology

Document technologies within Apollo refer to technologies whose primary
purpose is for the rendering and interaction with electronic documents.

PDF and HTML are the primary document technologies available within
Apollo.

PDF

PDF functionality is not included in Alpha 1 of Apollo, so we cannot go
into too much detail of how it is implemented. However, in general
Apollo applications, both Flash- and HTML-based, will be able to
leverage and interact with PDF content.

HTML

HTML was originally designed as a document technology, and today it
provides rich and robust control over content and text layout and
styling. HTML can be used as a document technology within
Apollo—both within an existing HTML application as well as
within a Flash-based application.

What Does An Apollo
Application Contain?

Now that we know what technologies are available to applications
running on top of the Apollo runtime (see Figure 1-1), let’s
look at how those technologies can be combined to build an Apollo
application.

Applications can consist of the following combinations of technologies:

• Flash only (including Flex)
• Flash-based with HTML content
• HTML/JavaScript only
• HTML/JavaScript-based with Flash content

All combinations can also leverage PDF content

Technology Integration
and Script Bridging

Because WebKit and the Flash Player are both included within the
runtime, they are integrated together on a very low level. For example,
when HTML is included within Flash content, it is actually rendered via
the Flash display pipeline, which, among other things, means that
anything that you can do to a bitmap within Flash (blur, rotate,
transform, etc.) can also be done to HTML.

This low-level integration also applies to the script engines within
Apollo (that run ActionScript and JavaScript). Apollo provides script
bridging between the two languages and environments, which makes the
following possible:

• JavaScript code to call ActionScript APIs
• ActionScript code to call JavaScript APIs
• ActionScript code to directly manipulate the HTML DOM
• Event registration both ways between JavaScript and
  ActionScript

Note that the script bridging is “pass by
reference.” So when passing an object instance from
ActionScript to JavaScript (or vice versa), changes to that instance in
one environment will affect the instance in the other environment.
Among other things, this makes it possible to maintain a reference to
HTML nodes from within ActionScript and modify them, or to register and
listen for events.

This low-level script bridging between the two environments makes it
very easy for developers to create applications that are a combination
of both HTML and Flash.

The end result of all of this is that if you are a web developer, then
you already have all of the skills necessary to build an Apollo
application.

Apollo Functionality

Apollo provides a rich set of programming APIs, as well as close
integration with the desktop that allows developers to build
applications that take advantage of the fact that they’re
running on the user’s desktop.

Apollo Programming APIs

In addition to all of the functionality and APIs already offered by the
Flash Player and WebKit engine, Apollo provides additional
functionality and APIs.

Apollo APIs will be exposed to both ActionScript and JavaScript.
Some of the new functionality includes, but is not limited to:

• Complete file I/O API
• Complete native windowing API
• Complete native menuing API
• Online/Offline APIs to detect when network connectivity has
  changed
• Data Caching and Syncing APIs to make it easier to develop
  applications that work on- and offline.
• Complete control over application chrome
• Local storage/settings APIs
• System notification APIs (that tie into OS-specific
  notification mechanisms)
• Application update APIs

Note that functionality may be implemented directly within the Apollo
runtime or on the framework layer (in Flex and JavaScript), or by using
a combination of both.

Apollo Desktop Integration

As discussed earlier, applications deployed via the browser cannot
always support the same user interactions as desktop applications. This
leads to applications that can be cumbersome for the user to interact
with, as they do not allow the type of application interactions with
which users are familiar.

Because an Apollo application is a desktop application, it is able to
provide the type of application interactions and experience that users
expect from an application. This functionality includes, but is not
limited to:

• Appropriate install/uninstall rituals
• Desktop install touchpoints (i.e., shortcuts, etc.)
• Rich drag-and-drop support:
• Between operating system and Apollo applications
• Between Apollo applications
• Between native applications and Apollo applications
• Rich clipboard support
• System notifications
• Native Icons
• Ability for applications to run in the background

Once installed, an Apollo application is just another native
application, which means that the operating system and users can
interact with it as it does with any other application. For example,
things such as OS-level application pre-fetching and switching work the
same with Apollo applications as they do with native applications.

The goal is that the end user doesn’t need to know they are
running an application that leverages Apollo. They should be able to
interact with an Apollo application in the same way that they interact
with any other application running on his desktop.

Apollo Development
Toolset

One of the reasons web applications have been successful is that they
allow developers to easily deploy applications that users can run
regardless of which OS they are on. Whether Mac, Windows, Linux,
Solaris, or cell phones, web applications provide reach.
 
However, success is based not only on cross-platform deployment, but
also on the cross-platform nature of the development environment. This
ensures that any developer can develop for—and
leverage—the technology. Neither the runtime nor the
development tools are tied to a specific OS.

The same is true of Apollo. Not only does Apollo provide the
cross-platform reach of web applications, but, just as importantly,
Apollo applications can be developed and packaged on virtually any
operating system.

In fact, Apollo itself does not have a compiler or specialized IDE.
Apollo applications just consist of web content, such as Flash and
HTML. Any tool that can edit an HTML or JavaScript file can also be
used to create an Apollo application.

A beta version of Flex Builder with Apollo support is included with the
Apollo Alpha. Because Apollo applications are built with existing web
technologies such as HTML and Flash, you can use the same tools that
you use to create browser-based content to create Apollo applications.
The Apollo SDK provides a number of free command-line tools that make
it possible to test, debug, and package Apollo applications with
virtually any web development and design tool.

The included command line tools are:

ADL – Allows
Apollo applications to be run without having to first install them
ADT -
Packages Apollo applications into distributable installation packages

While Adobe will be adding support to its own web development and
design tools for authoring Apollo content, they are not required. Using
the Apollo command-line tools, you can create an Apollo application
with any web development tool. You can use the same web development and
design tools that you are already using today.

Is Apollo the End of Web
Applications in the Browser?

So, by this point, you may be saying to yourself, “Gee,
Apollo sure sounds great! Why would anyone ever want to deploy an
application to the browser again? Is Apollo the end of web applications
within the browser?”

No.

Let me repeat that again: no.

Apollo solves most of the problems with deploying web applications via
the browser. However, there are still advantages to deploying
applications via the browser. The fact that there are so many web
applications despite the disadvantages discussed earlier is a testament
to the advantages of running within the browser. When those advantages
outweigh the disadvantages, developers will still deploy their
applications via the web browser.

But is it not necessarily an either/or question. Because Apollo
applications are built using web technologies, the application that you
deploy via the web browser can be quickly turned into an Apollo
application. You can have a web-based version that provides the
browser-based functionality, and then also have an Apollo-based version
that takes advantage of running on the desktop. Both versions could
leverage the same technologies, languages, and code base.

Apollo applications complement web applications in the browser. They do
not replace them.