Hey folks, there's nothing to see here right now. We spent three weeks experimenting with some new tech to see what we could build, and ended up with something that looks a bit like Rails, but entirely client-side and written in CoffeeScript.
The project driving the framework—an HTML5 mobile UI for Basecamp—has been on hold for about a month now, and we're still a ways off from having anything to show. It just isn't a priority for us right now. When (or if) we do have something we'll be sure to post about it on our blog.
Why not put it up on Github so we can help you finish it?
I'm currently a full time jQTouch developer, but I would much prefer a working environment like the one you're describing. I would certainly be willing to help develop it.
A couple of other smaller components are ripe for release, too. But the framework itself needs more work before it can be made public, and that work is directed by the needs of the application.
"It just isn't a priority for us right now"
I think the number of upvotes your getting demonstrates a sincere interest in such a framework. Really look forward to seeing more from this project.
Oh man, that would be great! All Basecamp iPhone apps are terrible... If you don't mind me asking, is the "hold" temporary, so will the project be resumed at some point?
For anyone interested in the idea of having multiple views in JavaScript (with their own URLs!), "Sammy" is a client-side JavaScript framework that's like Ruby's Sinatra: http://code.quirkey.com/sammy/
Indeed, what Ryan mentions sounds a bit like the ideas of Sammy, Backbone, and HTML5 local storage and offline caching blended together into a single tasty package.
If you'd like to browse an example Todos application (with annotated source code) that fuses Backbone.js and HTML5 local storage, here's a good place to start:
Edit: There seem to be a lot of folks who think that Sammy is the only way to get client-side routing -- nothing could be further from the truth. Sammy asks you to structure your entire app around an inappropriate faux-server-side API, and Sammy apps don't work correctly in Internet Explorer because Sammy doesn't use an iframe to set history in IE. Doing hashchange events yourself only takes about a page of code, and handling hash URLs is a relatively tiny portion of a client-side app:
That said, so many folks have asked for hashchange routing that perhaps it would be wise to build a Backbone plugin for it that smooths over the difference between pushState and hashchange...
After doing my experimentations with various mobile stuff, I tend to think I'd prefer investing time in something specific to mobile like this or sencha, and rely on a (Rails or other) app with JSON API for the site or back-end.
The target seems very mobile-specific. HTML5 is better supported on mobile, and the proposed model sometimes fits "desktop" web apps, and almost always mobile web apps.
There is a need for Coffeescript; programming Javascript directly leads to cross-browser issues. Something like Coffeescript can abstract away the per-browser style issues. (The main annoyance I notice is the behavior of trailing commas in lists. In Firefox, it works. In IE, it silently breaks. Coffeescript avoids this issue, since the code you type never sees the browser.)
I would personally find a Javascript->Javascript compiler more compelling, as it would mean I wouldn't need a new editor toolchain. But that wouldn't be as fun to blog about, becuase you would still have to type $("DOM") instead of $ DOM!
It's mildly annoying. Unlike Fortran, JS doesn't desperately need a preprocessor language. Adding another layer of "abstraction" and some syntactic sugar will make the process of developing good JS needlessly complex and fragile.
So "it turns out"[1] that Coffescript will make the process of developing good JS needlessly complex and fragile.
Would you care to elaborate?
From the coffeescript site :
> it compiles into clean JavaScript (the good parts) that can use existing JavaScript libraries seamlessly, and passes through JSLint without warnings. The compiled output is pretty-printed and quite readable.
When I used it, it cut ~25% of the LoC off my project. And it only took about an hour to learn because the documentation is so damn good. On the usefulness scale, I'd put it next to jQuery and backbone.js.
Googling around for Ratfor doesn't reveal the history -- mind sharing what happened with Ratfor in the end? From your tone, I'm assuming it ended in tears...
It was more a "plus ca change" thought, I actually liked the idea of CoffeeScript, although for my personal work I came to the conclusion that it's not that necessary.
Ratfor itself was mostly a stop-gap solution. Fortran 77 made things a bit easier, and a quite a few developers migrated to C.
I'm actually surprised that there aren't more preprocessors like CoffeeScript. Yes, there are a few languages that compile to JavaScript, but most of them don't aim for ease of integration of existing libraries. So it's just you, your language and the DOM. And if it's just there to supplement your usual backend language (Lisp, Java etc.), you probably don't care that much about node.js.
The trick with NIH syndrome is balancing, as with any other syndrom. Once you know about it, you can also detect situations where it's worthy rolling your own (GitHub's Resque is another successful example of that).
The project driving the framework—an HTML5 mobile UI for Basecamp—has been on hold for about a month now, and we're still a ways off from having anything to show. It just isn't a priority for us right now. When (or if) we do have something we'll be sure to post about it on our blog.