As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance.
Closed 10 years ago.
Ajax, Flex and Silverlight are a few ways to make more interactive web applications. What kinds of factors would you consider when deciding which to use for a new web application?
Does any one of them offer better cross-platform compatibility, performance, developer tools or community support?
Here's a quick rundown of each area (with lots of helpful links):
Cross-platform compatibility
Ajax works in any modern browser that can run JavaScript. Flex requires Flash or anything else that can handle SWFs but, once that's installed, it's a total freeride as far as compatibility. Silverlight is tricky and misunderstood so carefully consider your userbase before going with this MS foray into the rich web applications arms race. Also keep in mind that Silverlight is still in Beta, so it may become more widely used and installed in the future as it is developed.
Performance
I'm fearful of making too many statements about performance because it really depends on how much you are willing to optimize and the exact nature of your application. Also, some technology stacks are just never going to be very fast. Some people out there have been making comparisons, but again, it depends on a great many factors (even the version of the browser from which you are testing!). It's probably best to choose based on other factors and optimize once you've started to develop.
Developer tools
There are the "golden standard" dev tools for each of the three:
Ajax has basically unlimited options, depending on the rest of your technology and architecture choices. The big questions you're actually faced with are which libraries to rely upon, and Google has voiced a pretty well adopted answer with things like Web Toolkit. When you get right down to it, it's just XML and JavaScript, right?
Flex is from Adobe and, just like with Flash development, you'd better stick with their homegrown tools because--well--they're making the standards as they go along.
Microsoft has positioned Microsoft Expression Blend versions 2.0 and 2.5 for designing the UI of Silverlight 1.0 and 2 applications respectively. Visual Studio 2008 can be used to develop and debug Silverlight applications (from Wikipedia).
Community support
There is both official and unofficial community, corporate, and open-source support for all three options. Whichever you are already integrated with and which makes you feel most at home are very individual things, but I'll offer this advice: stick with what you know. If you are a MS developer and have MSDN as your homepage, you are probably going to think the Silverlight documentation and forums are really helpful. Flex has a very similar story; the forums are pretty good and if you're a Flash person already, you're going to be right at home with their documentation and user community.
On the other hand, Ajax is basically all over the place because you can implement so many different ways and use so many widely-varied libraries. Each library can have it's own forums to visit and mailing lists to lurk within for answers.
Once again, all three have corporate giants trying to foster their communities and to get the best support possible to the developers that will give them greater market share in the future.
The choice should in my opinion be mostly based on the nature of the application you'll be building (for example, if you need to manipulate vector graphics, Ajax is pretty much out), but here are some general guidelines:
Ubiquity
Ajax -- Supported by all modern browsers across platforms
Flex -- Runtime (Flash Player) has very wide installed base for Windows, Mac OS, Linux. Linux version was a bit buggy the last time I checked, though
Silverlight -- Runtime has quite low installed base (and no Linux support) at the moment
Choice of programming language
(Unordeded because of subjectivity, but note that Silverlight offers the most choice. Also note that the existing language experience of developers in your team should be taken into account.)
Silverlight: Any .NET language (C#, Visual Basic, IronPython(?), IronRuby(?)) (and XAML for UI definition)
Ajax: JavaScript (and XHTML for UI definition)
Flex: ActionScript 3 (and MXML for UI definition)
API Stability and compatibility
Flex -- Runtime is the same across platforms and browsers, more mature and stable at the moment than Silverlight
Silverlight -- Runtime is the same across platforms and browsers, less mature than Flex/Flash, v2.0 is still in beta
Ajax -- Compatibility problems across browsers (may be mitigated via Ajax libraries, though)
Web/Browser Integration
Ajax -- Content is native inside browser, based on standards: searchable by browser and search engine crawlers, subject to any standard UI practices the browser and operating system have established
Flex and Silverlight -- Content not native to browser (i.e. runs in its own little "sandbox/rectangle"): not necessarily subject to established UI practices for the given platform
Developer Tools
Ajax -- Your favorite code editor, browser and debugging toolkit for the browser
Flex -- Flex SDK is available for Windows, Mac OS and Linux for free and can be used with your favorite code editor. A Command-line debugger is included, but the Adobe-provided profiler is only available in the commercial Flex Builder IDE
Silverlight -- AFAIK, The SDK is available for Windows for free and can be used with your favorite .NET development tools
The web runtimes like Flex and Silverlight all offer enticing things, but come with two big costs:
They run only within a rectangle on the page, and don't interact with normal web widgets
They are only available to people who have that plug-in installed
Even the nearly-ubiquitous Flash isn't installed on every web browser, so by choosing to use a plug-in runtime you're excluding part of your audience.
In contrast, JavaScript (or Ajax) is available on pretty much every browser, and interacts better with normal web pages, but is a more primitive and restricting language. Using it for complex animations can be tricky, and you'll need to test your applications in more versions on more platforms to make sure it works.
Cross-platform compatibility isn't the issue it used to be, so the issue is this: Will you gain more in the features of a plug-in library than you'll lose in the audience you exclude?
My own answer has so far always been JavaScript/Ajax, but I'd re-evaluate that in any new project.
What is your audience: public web site or an intranet business app? Adoption rates are not relevant if you have a controlled audience who will install what is needed to run your app. However, if you need the largest possible audience to make your web startup a success then it may be critical.
What is your goal? Building something for the lowest cost? Learning new technology?
Can you leverage your existing skills? If you already know .NET then Silverlight gets a boost. Learning Flex may be interesting and useful, but is it more useful to you than more experience with .NET technologies? Remember to consider the opportunity cost of learning something totally new.
I don't see a clear technology winner at this point, and likely there won't be one for a long time, so the choice will come down to fairly subjective factors.
Other than what's already been mentioned here, another huge thing to consider is what your UI is going to be.
If you're going to be using a lot of advanced UI controls like trees, lists, tab controls, etc then consider the following:
JavaScript/HTML - No native support for anything beyond things basic drop down boxes, buttons, and text fields. If you want something like a tree control or tab control then you'll have to roll your own or find a third party library.
Adobe ActionScript - Native support for a wide array of advanced UI controls
Silverlight - 1.0 had very limited UI controls, but 2.0 will be adding many more and I'm sure we'll continue to see controls added in future releases.
Related
I am trying to evaluate whether Xamarin would be a good option for my project. The project is a large, complex app for Android and iOS with a lot of client-server communication. The user interface is a major focus and has to be really fast and smooth. Also, we plan to make large use of UX graphic effects (comparable to the Spotify app).
For now we are planning to go for two separate native apps using Java/Objective-C. However, the possibility of cross-platform code sharing would be very convenient for us of course.
Most opinions I've heard so far say that Xamarin - although far better than HTML5 apps - cannot match the UX of a native app. Also, I tested the following applications made with Xamarin (on Android):
Rdio
MarketWatch
Busch Gardens Discovery Guide
Sqor
Storyo
From my impression, none of them could quite match the speed and smoothness of a good native app.
If our focus is on a top notch user experience, would Xamarin really be a viable option? Can it really match a native UX? I am particularly looking for opinions from developers who have experience with large and complex cross-platform Xamarin applications. A few critical voices would be very helpful.
Thank you a lot!
I'm on the Rdio mobile development team, so I can make some personal reflections from that standpoint.
Xamarin allows you to write native applications in C#. Any slowness, jankiness, ugliness or bad-appiness usually has nothing to do with the Xamarin layer itself.
You save some time being able to share core business logic between your different clients, but you're still writing the UI from scratch, specific to the platform. You're just writing it in C#.
But while you save that time, you're spending it in other ways. All of those SDKs you want to use probably aren't compatible with Xamarin out of the box. You won't be pod install'ing that iOS framework, and you might be reinventing the wheel for handfuls of things. Xamarin takes advantage of the NuGet repo so you have a library of components that handle many of the things most people need (Analytics, Testing, Facebook SDK, JSON parsing, Database, etc etc) but it doesn't cover everything. And it certainly doesn't cover stuff that's out the day of an Apple or Google product announcement.
Any 3rd party code that you do want to import into your project will be done through writing custom bindings. While not usually difficult, it is time consuming. Xamarin has a team of people that specialize in assisting you in this. This fact speaks to the process being messy at times.
So while the slowness, jankiness, ugliness or bad-appiness probably isn't the fault of Xamarin, it might be the fault of you spending time in places you normally wouldn't, or not being able to take advantage of features you normally would. If that 3rd party partner SDK is giving you problems, your troubleshooting may take twice as long because there's a layer that you don't control.
UI is a wash. You're writing it from scratch anyway.
Business logic is shared. Depending on the app that might be a win if you architect your application to take advantage of it.
Compatibility / bleeding edge ability will be lacking. That might not matter to you at all, or you might be the person wanting to take advantage of that hot new API in the next OS release the day it's announced.
My personal thought, without knowing specifics, is if you want to build an application that you plan on being around years from now, and that will take advantage of the latest and greatest, I'd tell you to write natively for each platform. Unless you can really see huge gains in sharing that business logic the upfront gains are minimal. Or if you really like C#.
Xamarin uses native controls. So you design a fully native UI per platform. The users can't see that your App is made with Xamarin or Java/Objective-C.
There are sometimes performance issues in conjunction with the platform independent UI wrapper Xamarin.Forms. But you're not forced to use it. When you have still performance issues in your Xamarin.Android or Xamarin.iOS app then you produce them in your code.
There are benchmark results for Android apps comparing Xamarin.Android and Java apps: Does anyone have benchmarks (code & results) comparing performance of Android apps written in Xamarin C# and Java?
As you can see Xamarin's internal performance became better and better over the time.
Conclusion: Yes, you can write smooth native Apps using Xamarin.
I have to develop an application for Windows Mobile 6.x and version 7 and my skillset is .net and web based. Since version 7 is not backwards compatible I am thinking that I will end up with a lot of duplicate work.
I had a look at Rhomobile and Phonegap but the later does not support windows mobile 6.x and the former would require ruby knowledge.
I thought that I could have as much logic in class libraries as possible and then "include as link" the classes into projects so that they can be compiled for the specific platform. I feel like I may end up with a lot of #if PHONE7 style logic in that case and also the user interface code would have to be implemented for each platform.
What is the most efficient way to solve this problem?
If you can separate the UI from the logic you can begin to multi-target. The other consideration is the runtime of the phone (Silverlight 4?) and the phone sandboxing giving you limitations.
MVVM Light is multi-targeted and is open source, you could use this to review how they tackled it:
http://mvvmlight.codeplex.com/SourceControl/changeset/view/d546dc69fffd
In short though, your "add as link" idea is pretty much the basis of it.
The other question to ask is will the application fit on both platforms? The usage style from mobile to metro is a big change. People expect things to behave a certain way and look a certain way on metro. If you can confine this to the UI and move as much logic out as possible, then you should be ok. But there will come a point where it isn't worth sharing, it is simply easier to duplicate.
Sorry for a lack of detail in this answer, to be honest it is heavily app specific. I would personally start with separate applications and refactor as commonalities appear, using a combination of linked code files and hash defines.
A cute trick... you can always make the classes partial and have a partial code file for each hash define:
MyBusinessClass.CF.cs
MyBusinessClass.Silverlight.cs
What's your experience in using web technologies (HTML, XML, CSS, JavaScript) to implement part of the functionality of a GUI application? Pros and cons, please.
No servers, relational databases, AJAX, or cookies for session management, nor an existing webapp either, but rather a GUI app that uses web widgets (like Qt WebKit) to render and handle substantial parts of the UI, while taking advantage of a GUI framework to achieve an even richer interaction and better desktop integration.
I've already validated that the approach is possible using PyQt. Content can be rendered from the file system or from strings, and URL requests (images or clicks) can be captured and served by the form's event handlers. CSS and JavaScript are supported, perhaps with some limitations.
# ...
self.webView.page().setLinkDelegationPolicy(
QtWebKit.QWebPage.DelegateExternalLinks
)
#...
class TotiMainWindow(QtGui.QMainWindow):
def linkClicked(self, url):
pass # events arrive here
Note: This question is different from this one and this one made before, among other things because there is no requirement to use web technologies on the GUI, but there is the requirement that the application should work without a network connection available, and should integrate well with the default desktop over different platforms, without previous infrastructure requirements (no .NET, Java, browsers, or database servers).
Note: I posted a different version of this question on PMS but found very little experience with this approach there.
Closing Note
I just found most of the information I was looking for in a series of blog posts by André Pareis.
I think the largest advantage to using web markup like HTML/CSS and other web technologies is that desktop apps may very well have their days numbered.
As we speak, Google engineers are working on the Chromium OS, which essentially consists of a single GUI application... the browser...
Now, while nothing may never actually come of it, there is clearly a rising trend in the number of applications accessible through a web browser, accessible anywhere. It seems to me that this is the future of application development.
By using these technologies, this becomes one less headache you have to deal with when or if you determine that your app should be available as a web application.
Update: A few years ago, we developed an Agent Desktop for our call center that is essentially a local application that opens sockets to integrate with the phone system. The user interface the agents use is built with HTML, CSS, and JavaScript, and the experience is stunning. When we released our latest update in 2010 with a professional CSS redesign, our agents were all very impressed with not only how easy it was to interact but also how easy it was to use.
In the future we will port this 100% to the browser, but for now it needs to be a local application because of the COM integration with the phone system.
We did exactly this for a project back when Windows XP was new.
This gave my team several benefits:
A good-looking UI with relatively little effort
Easily change the style of the UI in a consistent manner using CSS
Relatively simple integration with C++ (invoking functions from the ui and vice versa)
The drawbacks we saw were:
Some not-so-good firewalls considered accessing internal resources (ie other html pages in the ui) to be a web request
Adding and accessing the needed resources could in some cases be a bit cumbersome
It was possible set properties in Internet Explorer that would prevent JS from running in the application
Note that some of Windows XP:s programs are using this approach.
This probably works best with small, more Wizard-like parts of the ui (which our ui consisted almost entirely of).
I have since then not really been involved in ui projects, so I cannot really tell you whether this approach is still valid... I know that MFC-based applications will let you use HTML-based dialogs though.
In a similar situation in 2005 I created a stand-alone webapp using XForms, CSS, JavaScript, XML and XML Schema for offline data retrieval and verification. With a good XForms -> HTML + JS transformer (Chiba) it did the job with no bug fixes after the initial release. It was used for 6-12 months (IIRC) by about a dozen engineers for a project gathering test data in the tunnel of the Large Hadron Collider. The biggest surprise of that project was just how much you get for free when going for a web platform, even for offline use. Highly recommended.
The major problem is that it reduces your development speed, or the quality of your user interface. A lot. Unless you're using Seaside, it is much faster to develop a desktop app.
There is quite some number of applications built on top of Mozilla platform. It isn't 100% web technology, as instead of HTML you use XML based XUL, but the rest is indeed web stack (JavaScript, CSS). The most successful of these it the OpenKomodo and it's commercial big brother Komodo IDE.
On the other hand, as far as Qt goes, the newest version 4.7 you can build GUI using QML language. Don't let the name mislead you, it's not markup, it acctually JavaScript with app-specific extensions.
I've been programming for over 15 years and started with .NET 5 years ago. We built our framework for windows data-oriented apps and it is a quite stable.
At this point, we are considering to make a new platform. But, I am a little bit confused with all these new technologies. We considered CAB and SmartClient technologies but there are also WPF, WinForms and Silverlight options in there.
It sounds like you just need a quick overview of the technologies you have outlined there.
WPF: Windows Presentation Foundation - a new graphical rendering system for building interfaces based on the XAML markup language.
WinForms: Windows Forms Applications - Visual Studio's classical drag-drop GUI.
Silverlight: a Web Application framework - usually used with WPF, very similar features
ASP.NET & MVC: ASP.NET is the web application framework used in conjunction with C# and VB.NET used and MVC stands for Model-View-Controller - a design pattern that has actually been around for ~30 years
Without knowing the true intricacies of your framework, what you need it to do, what limitations you have - I can't say X would be better than Y - especially seeing as WinForms and WPF is used for desktop applications, and Silverlight, ASP is used for web applications - unless you're thinking of linking these in with each other? You haven't given enough information in your question.
However, the best for investing in the following 5 years? The most recent and still in development technologies are WPF, Silverlight and the ASP.NET MVC - but nobody has a crystal ball to say whether these will still be alive, kicking, and technologically advanced in 5 years time.
We built our framework for windows data-oriented apps and it is a quite stable. At this point, we are considering to make a new platform.
For your specs, all choices might be very good :)
I know that, but what you think is the best for investing in the following, let's say, 5+ years?
If you have a choice ( meaning if you are just displaying data without all the fanciful animation, movie and game like stuff), use ASP.NET MVC.
I always advocate web applications over desktop ones, because web apps are hosted in a single place, and so they won't produce upgrade nightmare.
In terms of which one has the brightest future, well, none of us have crystal ball and we don't know what Microsoft will throw at us in a few years time. But you should orient your apps around your business, not around fashion. If you really ask me out of the so many, which one will still remain standing 5 years later, I would say ASP.NET MVC, because ASP.NET is a mature technology, and MVC is a tried-and-true design pattern that has been used by open source and non-MIcrosoft companies. The fact that it is widely accepted makes it less likely to fade out of fashion.
If you want to continue with creating desktop applications, then your two major choices are WinForms or WPF. WPF is the "hot" technology right now, but there are a few things to consider before choosing it.
WPF requires .Net 3.5 (maybe it's 3.0, but whatever). This is significant because many corporate customers are still using .Net 2.0, and may not authorize upgrading to 3.5 for some time (my former company still has some clients that haven't authorized moving from .Net 1.1 to 2.0 yet). Also, the installer for .Net 2.0 is 23 MB while the installer for .Net 3.5 is close to 200 MB, and the 3.5 installer appears to be buggy, failing occasionally with a helpful "SETUP Error" message.
Developing with WPF requires Visual Studio 2008. If you're currently using an older version, you'll have to upgrade.
This isn't completely true, but for the most part what WPF brings to the table is a much snazzier user interface. In my experience, corporate customers who are paying big bucks for custom software do not care in the slightest how the application looks - they only care that it functions well and that it ultimately saves them money.
Finally, if you already have experience developing desktop applications for Windows, then you will feel right at home going the WinForms route. Also, because of the Mono project, a WinForms app can (theoretically) work on other platforms like Mac, Linux and the iPhone, whereas this is not yet possible for WPF (I might be wrong about this, and I will happily correct myself if someone points it out).
I'm actually sort of torn about this advice, because I think WPF is very cool and powerful, and it may take off and become the only viable way of developing for Windows very soon. It is a risk, however.
Thank you all for answers !
We tested all platforms and we have at least one project finished using MVC, WPF, Silverlight, WinForms...
The main problem with WPF and Silverligth is missing of native ReportVIewer controls since we have a lot of reports (RDL) created.
Personally, I think also that MVC is best structured framework.
I really have positive thinking of MONO. But, I would rather decrease the price of the project so the clients can buy WINDOWS but to handle bugs in Mono-Linux option.
We decided to follow the Microsoft way.
I'm thinking of some kind hibrid framework (Webservice - SQL SERVER- ADO .NET Entity Framework) to be on hosted on server and hibrid CAB-SMARTCLIENT to be usen on clients.
I really miss the times where only one tool and technology was available like Levi's 501 :)
I'm a developer who builds mainly single page client side web applications where state in maintained on the client-side. Lately some of the applications have become very complex with very rich domain models on the client-side and increasingly complicated UI interactions.
As we've gone along we've implemented some very useful design patterns such as Passive View MVC, Observers, bindings, key-value observers (cocoa). I have recently got a lot of inspiration from the work of SproutCore and Cappuccino which are both JavaScript web frameworks inspired by Cocoa.
Obviously all of the problems that developers are having now in building complex web applications have been solved by desktop developers many moons ago. As few months ago all I knew about Cocoa was that is was some Apple thing, now it has had a big impact in the way I develop my web applications.
I was wondering if anyone who has more experience in building desktop GUI's than I, could point me any other frameworks out there which may also give me inspiration in terms of design patterns and structures to use for my JavaScript web applications?
I really don't care what languages or platform these frameworks reside in, as long as they can teach me something about good application design in general.
Fowlers GUI Architectures seems to be a reasonable survey done at a high level, I don't know how complete it is, however.
Have you taken a look at Cappuccino? It's a Javascript client-side framework, very heavily inspired by Cocoa. The Cappuccino creators even wrote their own Objective-C runtime in JavaScript so that Cappuccino apps can be written in Objective-J, an Objective-C-like syntax for JavaScript.