We Keep Coding

JavaScript GUI library which is usable from ClojureScript?

On one of my ongoing web development projects, we need to create a page (really a mini-application of its own) which will have an interactive, "Web 2.0" interface, more like a graphical desktop application than a typical web page.
I'm hoping to use a higher-level language which can compile down to JS for the client-side logic, rather than programming directly in JS, and ClojureScript is my first choice.
I'm looking for a JS GUI library which can be used from within ClojureScript, and which supports a style of programming similar to "desktop" GUI libraries like Swing, QT, wxWidgets, and so on. It should allow me to create popup <div> "windows", put widgets such as buttons/labels/etc. in them, attach event handlers to the widgets, and so on. Do you know of one which would work?

Clojurescript is still young - I don't think there are mature frameworks of that caliber (yet?). Having said that, I'm building a UI using the following components:
Twitter Bootstrap + JQuery for visuals and UI elements
Chas Emerick's Friend for server-side authentication
Compojure + Ring + Jetty for the server
Crate (Clojurescript Hiccup) for html/dom generating
Jayq for JQuery wrapping
Fetch and/or Shoreleave's remotes for client-server RPC
Shoreleave's pubsub mechanism or Clojure's add-watch for data binding
A good feature of Clojure and Clojurescript are that they favour simple functions on generic data structures. That's why you tend to see several smaller libraries that do one specific job, but work really well together, as opposed to giant monolithic frameworks with an object model that requires you to adopt one way of thinking completely.

You should consider Google Closure Library. It has all features you ask and is definitely usable from ClojuresScript, I've played around with it. See http://closure-library.googlecode.com/svn-history/r1430/docs/index.html

I think ExtJS would fit the best in your case, even though this is pure JS. This is the most advanced web GUI library at the moment and it makes possible to create web applications the way similar the one using the desktop languages and IDEs. Just check out their demos. I don't think there will be something similar for the ClojureScript in the foreseen future.

What combination of Javascript/Css/HTML frameworks and libraries do you recommend?

I'm looking to modernize my web development with some cool libraries, but there are so many tools that do different things, often with overlap, that it is hard to decide.
I'm not interested in one framework but more in a combination of various tools that work well together. I've looked at resets and normalizers, boilerplates, HTML frameworks, and grid framework, CSS authoring frameworks, CSS frameworks, Javascript frameworks, DOM manipulation, AJAX, resource-loading, feature detection and UI libraries, Javascript toolkits, server-side pre-processors and client-side pre-processors, and framework generators; It's beautiful chaos.
Also I'm looking forward to using SASS and Coffeescript or something similar so a combo including those would be nice.
Update:
I've decided on JQuery for DOM manipulation and I would also appreciate some kind of kickstarter/bootstrapper.

Depends for what. Use what you know best!
However, here are a few recommendations...
The Feather-Weights
For quick, no frills and no overthinking development, I'd recommend:
Underscore for general JavaScript Development
BackBone for client/server communication (using jQuery or Zepto.js for AJAX calls) and designing your models and event buses
Jasmine for JS testing
SASS for cooler CSS
HTML Frameworks, what's that? :)
CoffeeScript is indeed great, if you want to go down that road. If you are interested in CoffeeScript, you may want to look at Google's Dart as well, but it's fairly recent.
The Heavy-Weights (Batteries Included)
If you're building a rather complex web-application (speaking more along the lines of thousands of lines of code here), you need to take it up a notch and in that case I'd recommend you look at:
The Dojo Toolkit
Sencha's ExtJS
The whole Google Closure Tools stack
Dojo and Closure can be integrated in complex build systems and their compilers will allow for a good modularization of your codebase while keeping it easy to produce a strongly optimized deliverable. They also both contain their own module loading system, so you won't need additional libraries like Require.JS, and will only load the parts you explictly tell them too if you use a custom build. But be warned, they're definitely more hands-on and have a steeper learning curve.
The Google Closure Tools are definitely very comprehensive and are as batteries included as it gets, but they do not necessarily make development easy for you: they give you the power, but you need to know what you are doing.
Bootstrappers / Kickstarters
Some solutions offer to "kickstart" your project and to prepackage for you a collection of some of the above tools, sometimes offering you some customization, so you can easily get started and don't need to maintain your own kickstarter up to date with new versions:
Bootstrap
Foundation
HTML5Boilerplate
Kickstart
Skeleton
WARNING: I haven't really tried these kickstarters extensively myself.

Obviously, the answer here depends greatly on what you're building. If you're really getting into front-end engineering and making single-page web apps, I'd recommend reading Addy Osmani's blog. Recently, he's been blogging and giving a lot of talks about javascript MVC/MVP/MVVM systems, as well as scaling javascript apps.
If you want a laundry list, read this article where he gives his ideal stack:
Backbone.js for lightweight MV*
Require.js + AMD + RequireJS text add-on (to assist with external template management)
Backbone.js LayoutManager (if you require some more intelligent layout management)
jQuery for DOM manip.
Handlebars.js for templating, unless you're doing something simple, in which case, opt for Underscore's Micro-templating
r.js for handling script optimization
Jasmine + Jenkins for testing and CI
Node.js + Express (speaking of Node, Miller Medeiros has an excellent write-up on how to use it as a build script)
MongoDB as a noSQL data-store
This is surprisingly close to what I use myself and I can attest to the quality of these libraries and this as a whole-stack solution.

jQuery for JavaScript. Many great UI plugins for UI controls in addition to the code jQuery API that hides the browser differences in the DOM.

Angular JS - Awesome. MVC Framework, great for quick SPA's
jQuery - DOM manipulation - good for effects (fading, toggling etc.)
Don't rely on it too much though.
Angular JS, jQuery - Some argue they work well together.
Backbone.js
Ember.js
React.js - Uses JSX, not JS, but pretty much the same thing
<script type='text/jsx'>
//JSX
</script>

Pure JavaScript-based clients

I would like to know, how powerful/viable are JavaScript only clients based on say, GWT/gxt/vaadin, when compared to DHTML clients such as those made with wicket, tapestry, click etc?
My boss has insisted on using GXT (due to its nice colors and theme) on a project that will most likely become very big with lots of screens. I am against the idea of a javascript only client, especially when the javascript is generated from Java code. I have tried to advice him that we use something like wicket whereby we construct the screens with html but put in ajax where and when neccessary.
How viable is such a JavaScript client? I understand that JavaScript was intended for minor web page enhancements, and not all browsers, especially mobile devices have complete support for JavaScript.

Yes, it is viable for certain applications. Consider Gmail, Google Docs and Google Maps as typical applications where this works, and is probably the most feasible approach.
Some rich UI JavaScript frameworks, such as Ext JS also rely on this technique.

I've built javascript only web apps for ages.
First in SAP projects for big multinationals. And now on a new project:https://beebole-apps.com?demo
So yes it is powerful and viable.

Javascript-only webapp can be extremely powerful, and it's viable for certain applications, say, an Instant-Messenger webapp?
You mentioned that there are lots of screens in your web-app. One of the advantages from GWT/GXT is the fact that you can unit test your UI-layer with JUnit. This is an extra testing you can do on top of, say, Selenium. This is essential if you'd like to make UI testing a part of the continuous integration process, and, as the team grows, you'll definitely want to have tests around to make sure everything works (At least in theory.)
However, if what your boss meant to do is to build an in-house, custom Javascript engine using GWT's JavaScript Native Interface (Link), then I'm not sure...
Another advantage with GWT-like-engine over Wicket is that you can rely on HTML-code-gen to generate standard-compliant (In theory) HTML code. With framework like Wicket, it is hard to ensure every single developer on the team to author good HTML code - Especially when the team gets bigger.

Disclaimer: I'm a member of the Vaadin team.
Our Timeline demo is a good example of what can be achieved with Vaadin and GWT in client side, but I think all of the options presented in this discussions are viable given enough time.
Since you are going to start a big project you should build a simple proof-of-concept app with each of the relevant frameworks. If your PoC includes at least some of the more complex use cases you'll probably can make a pretty informed choice based on the experiences you get while building them.
I urge you to at least evaluate Vaadin. With it you write only server-side Java code and Vaadin will create a slick and professional browser UI for you. Client side can be easily extended using standard GWT (also pure Java), and there are no HTML templates, tag libraries or XML configuration involved at all. A Vaadin UI is fully Ajax'ed and lazy loading out of the box, and it easily integrates with any server side technologies, eg. Spring.
In addition to the development model advantages you get top-notch documentation, a bi-weekly update schedule, a very lively community filled with helpful experts, 100+ useful open source add-ons, and a 10 year old backing company with help on hand should you need it.

What kind of architecture is good for my JavaScript software?

I am writing a software for a website where people can manage the website content. The interface is similar to Windows with all those windows floating around, click-able components, type-able text fields, etc. The admin interface is written in JavaScript, HTML and CSS. I'm wondering what kind of architecture should I apply to my admin system before it's too late (I've written some minor code so far).
For example, for the PHP back-end I'm using MVC, but what should I use for the client-side JavaScript admin system?

If you were developing an GWT application, you would automatically have an architecture. For custom JS, with AJAX calls a two-tier architecture would be more than sufficient. However, I prefer going the event-driven way, where I have a layer full of listeners and another layer of handlers. I use closures for the most part and use custom page-ids. Works like a charm. I rarely separate the AJAX calls out, owing to the single-threaded aspect of JS.
P.S: I had to use the term layers for lack of better alternative.

You can use an MVC on the browser too.
But it may be overkill.
The minimum you can do is to separate the HTML(view) and the JS(logic) code.
We use a javascript templating engine (PURE) for our JS app, but there are plenty of other available on the web.

jQuery and jQuery UI are a good start.

Have a look at JavaScriptMVC. It's an MVC framework for front end JavaScript applications. Version 2 is jQuery based so you'll get the power of that fantastic library into the bargain

Should I use Google Web Toolkit for my new webapp?

I would like to create a database backed interactive AJAX webapp which has a custom (specific kind of events, editing) calendaring system. This would involve quite a lot of JavaScript and AJAX, and I thought about Google Web Toolkit for the interface and Ruby on Rails for server side.
Is Google Web Toolkit reliable and good? What hidden risks might be if I choose Google Web Toolkit? Can one easily combine it with Ruby on Rails on server side? Or should I try to use directly a JavaScript library like jQuery?
I have no experience in web development except some HTML, but I am an experienced programmer (c++, java, c#), and I would like to use only free tools for this project.

RoR is actually one of the things the GWT is made to work well with, as long as you're using REST properly. It's in the Google Web Toolkit Applications book, and you can see a demo from the book using this kind of idea here. That's not to say that you won't have any problems, but I think the support is definitely out there for it.
There's a neat project for making RoR/GWT easy that you can find here (MIT license). I haven't had a chance to try it out yet, but it looks like a good amount of thought has been put into it. One catch is that it looks like it hasn't been fully tested with 2.1 Rails yet, just 2.0, so you may run into a few (probably minor and fixable) errors.

If you are looking to integrate GWT with non-Java backends such as ROR, PHP etc., you should bear in mind that GWT 1.5 now supports JavaScript Overlay types. This feature lets you write classes that can be mapped over the top of native JavaScript objects to easily provide accessor methods for properties of those objects and other extended functionality.
See this link for more details:
JavaScript Overlay Types
So you could return JSON encoded data from your backend via AJAX calls, parse it into a JavaScript Object and then access the data through your GWT Java code using the overlay classes you've created. Or when you render your page you can render static config data as JavaScript Objects and read it in via this mechanism, rather than having to do an AJAX call to grab the data.

If you know JAVA, and have somewhere you can host it (like a tomcat or glassfish container) I would recommend that much more than using Ruby for the back end. The main reason is that then you can share all of your objects, and use the built in RPC mechanism. I've done this for quite a lot of our projects and it's a huge timesaver, not to mention that the code is less error prone, because you don't convert your java objects to anything and then back again.
I have linked my GWT with Rails before, using the to_json function in Rails and then reading the JSON in GWT. It's all supported, but it is far more annoying than just doing the back end in JAVA.
Of course if you have cheap hosting, then Java containers are pretty much out of the question, in which case I would think Rails would be the next best thing.

GWT is very high quality with a great community. However you do need to know CSS if you want to adjust the look of things (you will) - CSS can do a lot of the layout, just like regular web if you want it to. Libraries like GWT-ext or ExtGWT can help a bit as they have stunning "out of the box" looks but for a price (extra size to your app).

You can code everything in Java using GWT, and you can integrate existing 3rd party javascript libraries with it. It's very good. I've never used RoR much though, so can't say anything about that.

If you're experienced in Java but not in Javascript/CSS, then GWT is going to be a lifesaver (unless you want to learn them, of course). CSS has so many little fiddly details. It is not uncommon to spend half a day fixing a 2 pixel misalignment that only occurs in IE6.
I am not sure about how easy it would be to use ROR for the back end... It is possible, I am sure, since GWT ajax communication is just servlets. But they provide some really nice functionality for passing Java objects back and forth which you won't be able to utilize if your server isn't also using Java.

I wrote about some of the disadvantages of GWT recently. Mainly, the disadvantages are: long deployment cycle for changes to some parts of the application and a rather steep learning curve. As a seasoned Java programmer, the second should be less of a problem and if you use a seperate backend, the first is also mitigated (as a complete redeploy is primarily required when you change the 'server' part of the application).

GWT is a wonderful framework with lots of potential. Keep in mind that it's still quite new, though. There are some unresolved bugs that can really annoy you, and they usually require ugly workarounds to get past. The community is great but you'll probably end up with a few problems sooner or later that Google can't answer yet.
But hey, I say go for it. The potential for GWT is awesome, and I bet it's future will be bright.

You should definitely use GWT for a new project (it's pretty easy to use in an old project too).
I my experience it's very fast to learn and use. The compiled javascript code is much better than anything you could ever write by hand and it works fast too.
Another benefit is the ability to debug you're code (which is hell with javascript alone)

This blog has inputs from many experienced users of GWT and have some great discussion points. I personally have huge experience with varied UI Frameworks. I will add my two cents. Lets look at fundamental advantages and disadvantages of GWT
Fundamental Advantage
GWT takes the web layer programming to JAVA. So, the obvious advantages of Java start getting into play. It will provide Object Oriented programming. It will also provide great debugging and compile time checks. Since it generates HTML and Javascript, it will also have ability to hide some complexity within its generator.
Fundamental Disadvantage
The disadvantage starts from the same statement. GWT takes the web layer programming to JAVA. If you know JAVA, probably you will never look out for an alternative language to write your business logic. It's self sufficient and great. But when it comes to writing configurations for a JAVA application. We use property files, database, XML etc. We never store configurations in a JAVA class file. Think hard, why is that?
This is because configuration is a static data. It often require hierarchy. It is supposed to be readable. It never requires compilation. It doesn't require knowledge of JAVA programming language. In short, it is a different ball game. Now the question is, how it relates to our discussion?
Now, lets think about a web page. Do you think when we write a web page we write a business logic? Absolutely not. Web page is just a configuration. It is a configuration of hierarchical containers and fields. We need to write business logic for the data that will be captured from and displayed on the web page and not to create the web page itself.
Previous paragraph makes a very very strong statement. This will explain why HTML and XML based web pages are still the most popular ones. XML is the best in business to write configurations. A framework must allow a clear separation of web page from business logic (the goal of MVC framework). By doing this a web designer will be able to apply his skills of visualization and artistry to create brilliant looking web pages just by configuring XMLs and without being bothered about the intricacies of a programming language. Developers will be able to use their best in business JAVA for writing business logic.
Finally, lets talk about the repercussions in direct terms. GWT breaks this principal so it is bound to fail. The cost for developing GWT application will be very high because you will need multiskill programmers to write web pages. The required look and feel will be very hard to achieve. The turn around time of modifying the web page will be very high because of unnecessary compilation. And lastly, since you are writing web pages in JAVA it is very easy to corrupt it with business logic. Unknowingly you will introduce complexities that must be avoided.

You could also consider Grails ("Groovy on Rails") which gives you the benefits of a Rails framework and the use of the Java VM.

Our team recently asked the same question, and we chose to go with GWT, especially since the designer plugin made working with GWT more accessible to non-java experts on the team. Whoever makes this choice, just beware you DON'T use the GWT Designer plugin !! It has not been updated (in at least a year, apparently) to create a GWT application that is compatible with IE8.
Our team had almost completed our application layouts, which were working perfectly in Chrome, FF and Safari. Then they blew up in IE. IE 7 would load partial pages (but not composite includes), and IE8 was not even able to load up the application. It just hung.
The designer plugin has buttons that allow the user to add CellTable widgets that are not IE compatible (CellTable, DeckPanel, Horizontal Panel, Vertical Panel, among others). These will cause intense pain when the layouts have to be re-done in java without assistance from the designer.
Experienced GWT users love it, but the designer plugin will kill you.