Slashdot Mirror
Is Anyone Using the Google Web Toolkit?
eldavojohn writes "After seeing some applications from Google and participating in the Google Codejam (which seems to be built using the GWT), I kind of expected to see websites spring up left and right based off the GWT. Well, it's been a year and a half since they open sourced it and I have to admit that I am more than a little disappointed by its low profile in the UI community. I've been trolling their blog and have seen a few books out on it. But the one thing I'm not seeing is its use outside of Google. I've worked through the examples and tutorials at home and though I've been impressed with the speed, I am disturbed by the actual result — a whole ton of generated Javascript. But this is the first UI technology I've found where I can write in the native language of the server (Java) to generate and unit-test the UI code. Aside from Google's use and the games of Ryan Dewsbury like KDice & GPokr, does anyone know of major sites using the GWT? If you don't and you've used it yourself, why isn't it taking off? Is it too immature? Is it a solution to a problem that already has too many solutions? Is it fundamentally lacking in some way?"
To me, the biggest problem is abolutely no fallback to non-javascript browsers. I'm not so much worried about users, but search engine bots won't be able to spider me and drive traffic to me.
-Bucky
...just not as much as you might expect. Part of the issue is that it's designed for when you develop your application from scratch. Generally with the intent of developing a desktop replacement webapp. The only problem is that not many companies are investing in such apps. They're investing in using DHTML/AJAX to make their sites more interactive rather than replacing the HTML interface outright. In that situation, GWT is not the ideal solution. (e.g. For quite a while, you couldn't even have more than one widget per page!) It works though, so you'll find it pop up here and there.
On a personal level, I'd rather see the effort spent learning GWT applied to learning Javascript and the web technologies instead. There are a lot of frameworks out there, but none of them are actually needed in 90% of the cases. What we actually need are programmers who know how to write maintainable and highly interactive Javascript components for their sites. Such knowledge allows them to get the job done faster than mucking about with Yet Another Framework(TM) designed to take a cannon to the problem of killing a fly.
Javascript + Nintendo DSi = DSiCade
I'm a long time web developer but I've never even cracked open the box on GWT, so take this with a grain of salt.
The idea of depending on generated javascript scares me. I'm against writing Javascript in Java, Ruby, Python or anything else. Javascript is just too much of a beast to debug to leave everything up to an opaque framework, and I want to be able to get my hands dirty. I like the smaller and more traditional open-source style frameworks. Prototype, jQuery, MooTools, even Dojo just scare me a lot less.
It could be totally irrational, and it also could be the fact that I tend to build web applications that need minimal state and pretty basic AJAX interactions. Nothing anywhere near as dense as, say, Gmail. If the right project came along I'd definitely give it a more serious look.
There's a lot of corporative GWT-apps because it's, probably, the best toolkit for rich-client web applications.
However, it's not used much in the public web because most sites just don't need that kind of user interfaces.
Also, GWT is incompatible with web spiders.
I spent a month using it to build something in April 2007, and it was a big ol' pain. Getting something that roughly worked wasn't too bad, but there were a number of bugs and quirks that cost me time and headache.
Also, their rendering components generated an awful lot of hard to manage HTML and CSS. There were several display issues I never got quite right, and when I asked a front-end whiz to help me out, he had some very unkind things to say about the generated code.
My end impression was that it was about 0.7 in quality, needed a lot of polishing, and was really only useful for Java people who didn't want to understand what was actually going on in the browser. Were I to do that project over again, I'd instead just use something like JQuery, and learn how to do JavaScript properly, rather than hoping a framework would save me from my ignorance.
Because most web developers don't know Java. They know XHTML, CSS and Javascript. Some know PHP. Most sure as hell don't understand how Java can be used to make a site and don't want to bother learning a whole new "language" (as it seems) to build one. So they dely on jQuery and Dojo which is pure Javascript that they understand.
I've been on a project using GWT in 2007, been quite successful. If you want to see an example that is public run over to Parlays.com, they have a Flex and a GWT version.
If you want to write clean code check out my blog on TDD with GWT: http://is.gd/1156.
With the 1.5 release they did some very promising improvements.
So you're right if you say it is not mainstream, but to say nobody is using it is exaggerating. Just be patient, GWT will continue to grow.
Show a man some news, distract him for an hour. Show a man some mod points, distract him for the rest of his life.
We have used it for a fairly big internal application for one of our clients. Given we wanted ajax rather than a typical rich client, the main advantage of GWT was that we could program in the same language end-to-end.
We managed to avoid a lot of boilerplate code by using the same data class definitions (POJO's) in the server and client. So an object might be created by hibernate from a database record, copied to the client, displayed and edited, copied back to the server, manipulated there and finally updated in the database via hibernate.
The main omission in GWT is a good framework for binding data to UI elements. Because there is no introspection available in the GWT client environment, it is hard to do this in a generic way. We solved the basic problem by generating class and property descriptors during the usual hibernate code generation step. We then created a UI-POJO binding framework that picks up and uses these descriptors. Again avoiding a lot of boilerplate.
Our code for all this is here: http://code.google.com/p/gwt-hibernate/
I'd say GWT worked out pretty well.
No, the cool, unique properties of a web app are pretty much entirely the user experience -- the fact that there's nothing to download, and no updates to manage.
I develop a rich client application for internal corporate use, and I find that casual users really miss web-style navigation. I get a lot of requests that are essentially requests to simulate a web experience by providing a bunch of screens that users can click through to find the information they want, instead of using traditional (perhaps formerly traditional?) GUI ways of exposing functionality.
Also, these days, mashups and Greasemonkey scripts really magnify the value of web applications. Deprecating a web application in a big company can be nearly impossible because you find out that there are a bunch of business processes that depend on mashups and fancy Greasemonkey scripts that have been hacked together (usually by interns, IT guys, and other random people) and that provide substantial business value.
Sorry, just have to ask: What do you recommend instead of jsf? This comes from a guy who has to choose a java web stack for an upcoming project, that'll have to scale. No need for a flame war, just asking for a short answer here :)
The wacky Javascript that GWT generates is fine if your a Googler and understand and can fix issues that might occur but there many other cleaner solutions out there that mere, non kool-aid drinking, mortals can much more effectively work with.
Perhaps I didn't say what I was trying to say ;)
:/
What I was trying to get at is that, I've mostly mastered the concepts, and now I want to try everything that I can. Almost everything that I do is for the challenge/thrill of it. Yet, at the same time, I desire that what I hack around with should become a tool in my toolbox. It's almost like writing code that never gets used; it just feels wrong to pour your blood, sweat and tears in to a project that never gets used.
I guess it's in my blood, my great grandfather was an architect on the Hoover Dam and Chrysler building, my grandfather a civil engineer (apparently, he used to slam trains together in White Sands, NM. with explosives on them to see how much shock would set them off), and my uncle owns a software company. It's like I have to find the biggest, baddest problem that I can, and then I must own it! And not just own it, but own it from three angles at once. Then, I put it down and almost never pick it up again unless someone asks me how I would attack that problem. I intern for a small local company where I maintain a Linux server, an old Access database, and write some bash scripts... and play with new 'shiny objects' when I've completely my assigned work. (my boss is a Comp. Sci. geek from back in the day - I can usually get him to let me geek out with stuff since he enjoys it as well, and on occasion, it's worked out well for the company) But when I get home, I want to build something! The more elaborate, the better!
With SDKs, it's the same thing. I'm fairly sure that between kernel hacking and the release of java.util.concurrent with Java 6, (or maybe they extended it...) I failed a course or two. To be honest, while immature, I'd probably say it was worth it for the hours of fun it provided to fully grok the Java threading model or the interrupt routing for an SMP kernel during a critical section of code. The icing on the cake was when I got to hack around more the next semester while we were covering those topics. Or being able to speak somewhat intelligently on the use of 'goto' in class after reading about it in Code Complete. Sure, the extra pay off is nice, but it's what I would have been doing regardless.
Maybe I'm sick, but I love programming languages. Last semester I was required to do a bit of LISP and PROLOG for AI class, ADA95 for procedural languages, and took up PERL for fun. They're all really fun to hack around in (except I can't really find a great use for PROLOG), and I'm always on the lookout for a chance to break out ADA again. Take it as you will, I also enjoy using recursion. It's too elegant to not admire.
So, for mine own part, I've found that solving problems is relatively simple, just break it down in to atomic sub-problems, start with the most trivial, and divide and conquer each layer above it; if the solution makes sense, it works. I just want a few more hammers to try out on other problems. Unfortunately, mot problems dictate the most correct hammer by the time you've broken them down to isolated cases, and usually the most boring hammer is the one you should employ.
If I mod you up, it doesn't necessarily mean I agree with what you've said, sorry.
Apart from a few niche Web 2.0 sites, most websites are still built using tried and tested backend tech, and laid out using HTML, CSS and some graphics. GWT is pretty much doing everything using Javascript and a little bit on the server side serving xml/json. Not everyone needs AJAX. Most sites need to be able to work without it (for accessibility, backwards cinpatibility and non-javascript visitors), so unless its capable of adding really useful features (cases of which are few and far between) AJAX and GWT are just not necessary. Its nice if you can have it, but its a luxury you don't actually require for a usable website / web application.
Caesar si viveret, ad remum dareris.
Because I needed a website with a high level of interaction. The client asked for enabling disabling of various things on a widget, some bells and whistles,but nothing fancy. In the beginning I wrote the code for this using javascript, hand coded the whole thing. But change requests, and much more important than that, browser compatibility problems cost me a lot of time. GWT fixed this aspect. Mostly compatible with all major browsers, and being much more experienced in Java than in js, I became more productive.
However, I should have limited my implementation to a single widget, and that was my mistake when using GWT. Use a plain jsp page, attach the widget to a div, and be done with it. Instead I've built the whole thing on GWT, and later fell in a position where I can not easily add very simple stuff. The usual GWT app is one single js chunk, which navigates to different pages by hiding and showing things on a page. This requires a little getting used to, and I've implemented more flexible things like pulling html via remote calls etc. But in general, mixing GWT with a more server side oriented technology (asp.net, jsp, jsf etc...) looks like a better approach now. But when you have to build a slightly complex interface where there are trees, enabled disable compoenents, users adding, removing things to a list etc, GWT serves well. I guess the secret is in the balance, just use it at the necessary level, no more. I could have used Flash, but that'd be a total pain for multiple reasons. (a lot of reasons actually)
As has been previously stated, there are multiple solutions to an ever-changing problem - rich internet applications inside the browser - but I think Google have come up with a clever solution in GWT, not just as a front-end UI framework but also in terms of a solution to scalability, maintainability and deployment.
To make anything work in a real environment, a GWT front-end is going to have to talk to a back-end somewhere. If Java really is your thing, you get RPC talking to Tomcat out of the box, or in my case I chose a REST solution. As a tool to complement your other frameworks, GWT never gets in your way, allows you to work at the DOM level where necessary, and fits in well alongside other 3rd party solutions.
Mozilla have their solution with XUL although you are stuck inside Gecko, and who knows what is going on with Microsoft and XAML, and of course there is Flash with its install base. However, with GWT, Google are producing cross-browser compatible output across all major browsers and they picked Java to do that. Using an established language does have benefits (existing frameworks/expertise/unit testing/debugging) but it is how they use that language that makes it clever.
http://riflethru.com/ is an interface for searching eBay that I developed using GWT, as is the iPhone version. When first picking up the toolkit, the article I had read stated "they have taken HTML, Javascript and CSS and turned it into byte-code", and upon further inspection, it turned out to be true. In my experience, GWT is versatile, capable and on the march.
The web is reaching everywhere so a solid HTML solution like theirs sits well in different environments and devices, but Java is certainly not for everyone. My guess is that this has little to do with the limited adoption, of which GWT does not appear to be suffering as a result. In addition, the publicity from Google I/O will fuel the fire, no doubt, and I see the YouTube figures on their presentations are numbering in the 1000's, although I had to watch Deferred Binding 3 times myself.
Theoretically...
Number of Java developers > Number of Flash developers > Number of Silverlight developers
Number of Javascript capable browsers > Number of Flash enabled browsers > Number of Silverlight enabled browsers.
Which is a good idea, since Google has created a framework in a language that most developers are familiar with, for a platform that just about all web browsers support out of the box.
However...
Number of PHP hosting sites > Number of ASP hosting sites > Number of Ruby hosting sites > Number of Tomcat hosting sites
Which is probably one of the reasons why it's not doing so well.
GWT-RPC is excellent. It allows me to use the same data objects on client and server, and debug both from the same IDE. But it requires a Tomcat server.
Now if GWT is able to compile the server portion to easily deployable PHP code, this could lead to somewhere interesting.
D'oh. I'm a Googler but seriously, nobody is meant to understand compiled JavaScript. Good thing then, that GWT provides a mode in which it does not minify the code, so if something does go wrong in the generated JS you can track it down in the human readable version.
There are a lot of GWT myths floating around here. I don't work on it but I did adopt it for a project of mine recently, and it's worked out pretty well so far. Myths I saw reading this thread, in no particular order:
I suspect you'll see GWT used more and more in future. It's such a radically different approach it takes some getting used to, and whilst it does support evolving an existing codebase towards it, it's still easier if you start from scratch. Still, I do believe that GWT is one of the nicest open source things Google has released for a long time. It's very well thought out and is designed to be very efficient, which is important for any non-trivial web app.
We did a QA interface to one of our projects so we could test the backend without going through the rich client. It was a great experience and really easy to get working. The javascript/java backend communication was ultra-simple to get working.
But...
If we want to really easily create something that looks great we'll be sticking with more traditional approaches either using Flash/Flex or AJax with a standard JS library. Having a designer skin a GWT app is harder than those approaches.
I've been reading a lot about DWR and that plus a UI library will probably give you most of the benefits of GWT. Have to give it a try soon.
Well, I work for a large TV company in France and the fact that Java code is easily maintainable and that it is our language of choice for server-side stuff makes GWT the ideal choice for GUIs. We have deployed 5 large web applications built with GWT and the users are really pleased with the experience. That said, we do control the deployment environment up to a certain point (everybody has JS enabled on its workstation and the workstations are fairly recent) and we don't mind about spiders not being able to crawl the sites.
It doesn't have any of the good widgets that exist on most of the google pages that make them great, at least the last time I looked. For instance, there's no pop-up calendar date picker like Google Calendar uses. There's no form field completion popup with scrollbar. So if you want to do anything good, you have to roll your own anyway, or use something like YUI.
If you are going to end up using YUI anyway, it does everything GWT does and more, so you might as well simplify and limit your third party library complexity to one fewer library.
Also, when something doesn't work in GWT, you usually end up debugging the javascript anyway. It's a lot easier to navigate code you wrote than generated code.
It really doesn't buy you much over JS-only toolkits, IMHO. Unless all you know is Java and you want to make a very very simple website.
I've noticed the same thing.
I'm a partner in a mid-sized software dev firm. We're a job shop that is slowly bringing our own product to market.
Anyhoo, we wrote a route costing app for a certain airline (hint: It wasn't a foreign airline) about 5 years ago and we've upgraded the app about once a year since.
Last year we were asked how much it would cost to re-write as a web app.
It's not a broad app in terms if its function. But it's got a set of algorithms that are very tightly wound and the best guess of my web guys is that writing it in an interpreted language would slow it down considerably. (We're primarily a Python, PHP & Ruby shop when it comes to web dev.)
I told them and asked them why they were unhappy with the current Citrix nFuse (or wevertf it's called this year) deployment.
They weren't, they told me.
Long story shorter, they just wanted a web-style UI.
And they wanted it bad. So bad, that we re-worked every screen in the application.
We made it look like web browser.
There's an address bar, screens can be book-marked, back and forward buttons, all buttons in the app have been re-styled to hide the button border and put a blue underline on the button text. Etc Etc Etc.
Luckily it's not MDI or we'd have been creating a tab interface :)
honestly, that re-design probably cost $25,000.
No functionality was added. Just a different UI.
And this is why i think "cry me a river" whenever airline CEO's are crying to congress about not being able to stay afloat.
So you write all your software in binary?
How is writing in Java and compiling to JavaScript any different than writing in Java and compiling to bytecode?
* Keyboard support to Menus and TabBars
* Added ARIA roles/states to MenuBar/MenuItem, Tree/TreeItem, TabBar/TabPanel, CustomButton/ToggleButton/PushButton
* Screen readers are now able to identify and speak the content of these widgets
* Improved tab navigation
* New API to set ARIA roles/states on Elements (still experimental)