Building a leaner XulRunner for HTML/JavaScript usage? - html

Is there a way to make a leaner HTML/Javascript -oriented build/package of the Mozilla XulRunner components, excluding some of the heavier XUL-platform features?
Background:
We are building a cross-platform application with a HTML/JavaScript GUI. The host app is running on .NET/Mono, and the GUI is rendered using Mozilla XulRunner via the GeckoFx wrapper. Everything is working great.
We have one issue: the XulRunner (9.0.x) components weighing at 28MB make the application installer a bit larger than we'd like. The XulRunner does not need to be deployed separately if the user already has Firefox installed, but we would want to reduce the footprint for the users who don't.
We've already excluded all XulRunner executables from the official release build to get the package down to its current size.
It occurs to me that we are only using a small subset of the XulRunner platform. We do not require any of the XUL features, only HTML, JavaScript and CSS. This leads me to think that we might be able to make a custom build which excludes some of the parts we don't need, but I'm not well-versed in the Mozilla build arcana.
Any help would be greatly appreciated!

You will typically use build options like --disable-tests (don't build those test executables) and --enable-optimize/--disable-debug. Also, you should be able to use --disable-xul option. However, you need to be aware that options like --disable-xul create builds that aren't quite as well tested as the official XULRunner. Other options that should be ok in a simple application:
--disable-jsd Disable JavaScript debug library
--disable-plugins Disable plugins support
--disable-view-source Disable view source support
--disable-accessibility Disable accessibility support
--disable-printing Disable printing support
--disable-crashreporter Disable breakpad crash reporting
--disable-mathml Disable MathML support
--disable-installer Disable building of installer
--disable-updater Disable building of updater
--disable-activex Disable building of ActiveX control
--disable-activex-scripting Disable building of ActiveX scripting support (win32)
--disable-feeds Disable feed handling and processing components
You can find other options by searching for MOZ_ARG_DISABLE_BOOL in configure.in.

Related

Using Electron vs. Offline HTML5 for an offline application

When looking up Electron and Offline HTML5, I have found it difficult to make a decision between which one to use for a project.
Assuming that a user will have to go to a certain website to download the Electron application, and would have to go to the same website to get the Offline HTML5 loaded, what are the pros and cons between using one over the other?
Some that I could think of:
Offline HTML5 can be updated without the user consciously updating their application by just making the user go to the online page again.
Electron would eliminate the need to code around multiple browser/browser version dependencies and quirks
It really depends on your exact requirements. The following is a list of everything that I came up with:
electron supports module system (i.e. require) both in main and renderer processes.
electron provides you access to OS APIs (e.g. fs). Without such many node modules will not work in the JS runtime of the browser (e.g. ip).
updating your app with electron is as easy as sending an http request. (or even better as described here)
an HTML 5 offline app requires a browser and the user might just give it IE6.
electron integrates with native desktop environment (see dialog, power-save-blocker, shell or even app for some examples)
electron enhances some of HTML5 APIs such as file API
electron lets you modify the default behavior of underlying chromium. For example you could intercept all URLs with file scheme and modify them on the fly*.
In short if you want your app to have a good native integration and act deterministic (i.e. no browser quirks) I suggest choosing electron.
*electron-jade for example takes use of protocol API to compile all files ending with .jade on the fly without the need to prior compilation.
DISCLAIMER: I am the developer of electron-jade.

What cross-browser technology do you use in your web applications to manipulate on client machines?

(I have a problem with Google Chrome improvements that will drop support for my current solutions.)
I work on project where I move desktop system to an Intranet web application.
The crucial requirements are:
to move desktop system to a web application
to reproduce every single functionality from the desktop system in the webapp
While 95% of work requires creating casual web application, there is one thing which is non-standard to handle: my application must perform some actions on the client computer. These includes:
connecting with libraries
launching desktop apps
file manipulation in background
The example scenario is to integrate my system with some machine in the lab. I have to integrate my web application with drivers on client computer via DLL (desktop app did this, so my app have to do this as well).
Theoretically scenarios of the desktop actions may vary and I just want to implement some interface that will handle all the "client-machine" job the desktop app has done, so there is no need to work on every single scenario (but of course every scenario should be tested).
My solution was Java Applet. It worked. But then Google Chrome decided to drop support for NPAPI plugins, so in September'15 Java plugin (and my applet) won't be supported (http://blog.chromium.org/2014/11/the-final-countdown-for-npapi.html).
So my another solution is Java Web Start. It works. But now Google Chrome decided to drop support for background operations of external protocols (https://code.google.com/p/chromium/issues/detail?id=348640), so from Google Chrome 45 my Java Web Start solution won't be supported.
(Both above solutions work on Firefox and IE.)
The question:
What other technology can I use to interact with a client machine from my web app?
Other remarks:
I am reluctunt to write my own PPAPI plugin or Chrome Extension - I prefer one solution working on all major browsers.
I know that StackOverflow community does not like discussions about technologies, so please focus on describing possible solution to my problem.
We struggled with a similar problem as we need to connect/access electronic devices over JNI->DLL. The only technology where this is currently possible are applets. Period. (And even that is tricky since certain combinations of browsers/java versions/operating systems do not work or have problems, but this is another story...)
There are web technologies like HTLM5, JScript which can replace some functionalities of applets however in certain scenarios (like yours) there is no current alternative available - and you named some of those:
connecting with libraries like *.dll, *.so etc.
file manipulations
launching applications
And doing that across browsers and operating systems!
Solutions?
Tell your users that certain browsers can't be used (like Chrome and
Opera Next)
Write individual plugins for each browser (which probably is beyond your budget ;-)
Did you consider writing standalone application(s) in form of an executable file? The user must download and run it however e.g. java or plugins also need to be installed. But then there is the security aspect of that (downloading an and executing an executable file) - certainly not an easy decision
Have a look at FireBreath 2 - (just read about it in some posts, however didn't try it)
There are lots of discussions on SO to this topic so take a read:
alternative technologies to replace applets
applet alternative launch from browser
alternatives to java applet to launch microsoft office applications
alternative-java-applet-network-drive-access
what are the alternatives for java applet to launch client programs using chrome
alternative of npapi plugins for flash java applet
python alternative to java applet
npapi alternative for live file editing
... and many, many more!

Clickonce: Appropriate for web site that is only supported in Chrome?

We are looking for an easy way for our users to download and install a small client widget (it is a windows app) used with our (much larger) web app.
Our web app is currently only supported in Chrome.
Our concern is that the clickonce support piece may be missing from most Chrome installs.
The questions:
-- Is clickonce really going to be useful in this situation?
-- What is the best option for "one click" download-and-install for Chrome on windows?
ClickOnce uses NPAPI which is being removed from Chrome. You will need to look at NaCl, Native Messaging or PPAPI instead of ClickOnce. There used to be browser extensions that you could use, however they have been removed from the store (or no longer work).
Using Chrome should not matter, so yes Click once will be useful in this situation.
The only thing you need to worry about is your clients using Windows and the appropriate version of the .net framework. I hope this helps.

Is a NPAPI plugin what I am looking for?

I need to develop a mini ClickOnce plugin for Firefox, Chrome, Operah and Safari (preferably, one that works on all of the above).
My plugin needs to be able to run an application in a one-click fashion. In other words, by going to a specific URL, my application will run on the target machine (no prompts / dialogs shown). Unfortunately, ClickOnce is only available with IE, and I am unhappy with the ClickOnce plugins that currently exist. I would like to develop a tailored ClickOnce plugin that suits my specific needs.
I read about FireBreath, which sounded like a perfect solution for me. I also read about the difference between an extension and a plugin and I think a plugin is what I am looking for. I just wanted to make sure with you guys that I am on the right path. Is what I want to achieve possible with FireBreath / NPAPI plugins? Can a plugin download an executable and then run it on a target machine?
Yes, NPAPI will allow you to do what you want. You can run arbitrary native code using an NPAPI plugin, and control it via JavaScript on the web page side. Users will have to download and install your plugin though, so it only makes sense if this is something users will use more than once.
You'll obviously have to be very, very careful about security though. If you make a plugin whose sole purpose is to download and run code without user interaction then you'll need to be absolutely certain that there's no way for a malicious page to use your plugin.

complete client side application using HTML 5

Is it possible to create a complete client side application using HTML 5 (it will embed chat clients, stock ticker) + JavaScript, and then bundle is with underlining Firefox 3.5 engine, so that user can install it as a desktop app on Linux, and then it runs in its own window using firefox 3.5 engine. The app will store everything in local storage - like url, usernames etc, and will not interact with any local server.
It would pure HTML5 + JavaScript based client web app which would be packaged, and can be installed or launched from startup scripts.
Or instead of using firefox 3.5 engine, we can use Mozilla Prism to convert it to web app, and then bundle it.
Well, you should take a look at xulrunner (https://developer.mozilla.org/En/XULRunner). This is an engine to build portable applications... like firefox. You are getting full Gecko engine + full access to the environment, like a plain application. (XUL is something similar to HTML -- it is still XML-based UI description -- but it is better suited to develop desktop applications -- using native/almost native UI).
It sounds like what you're trying to do can be done with Adobe AIR (though there is no HTML5 support yet). Adobe AIR apps can be written with just HTML and JavaScript, though AIR uses the WebKit engine, not FireFox's.
The app part is surely doable (and will work just about everywhere, not just Linux), the bundling appears to be the more problematic part -- not every distro of Linux will like the specific way you built the bundled Firefox, not to mention that you're cutting yourself off from just about any non-Linux box on the planet (and every linux box with a non-x86 compatible processor, etc, etc)... all for the sake of that bundling. Can't you stop at the prism aka webrunner part, without bundling?
I'm almost sure I've seen something like this done with Qt Webkit. It was only a few lines of C++ to create the window with the web browser control and load a local page. The page created the UI of the application. I can't find the link though.
I second what liori said about using XULRunner. You can deploy your HTML application easily, and, more importantly, when you run into limitations of HTML 5 or its level of support in Gecko, you can work around it by using XUL APIs.
(AFAIK, Firefox 3.5 doesn't support any socket APIs in HTML content, and it doesn't support offline storage in web apps, only the localStorage part).