Thoughts On Sproutcore and App Dev In General

I’ve been intentionally learning how to develop using Sproutcore for about a month now, so I’ve been more aware of the buzz about it, and I have to say that most things I read are missing the point about the value promise of Sproutcore.

The impetus behind my learning is that it is an application development framework for the HTML5/JavaScript platform, versus the Flash Platform that I’m used to developing for. The key here is that it is an application development framework. It is not another DOM access library, widget set, or MVC, and it shouldn’t even be compared to them. It’s not a “Flash-killer“. It’s not even a “Flex-killer” per se, although in my mind it is an alternative in part.

Flash is just a display technology. Flex boils down to just a widget set and bindings library. Intrinsically, Flex lacks a good application development pattern. That’s partially why Seppuku is a viable alternative to maintaining the average Flex app built by someone else. There’s no built-in MVC, or unit testing strategy or IoC methodology.

I realize that Adobe’s building in support for FlexUnit4 in Flash Builder 4 / Flex 4 SDK which is wonderful. Michael Labriola et al are doing a fantastic work there that is critical for the Flex project. These necessessities have been the mother of invention with MVC frameworks like Cairgorm, Mate, FlightFramework, and HydraMVC for MVC. Lack of IoC has also inspired Mate, and Flight, as well as the excellent Swiz framework, and the DI part of HydraFramework. And, don’t even get me started on server interaction complexities.

I say all that to illustrate that the Flex development stack really fundamentally consists of Flash -> Flex (widget set, bindings, themes) -> MVC -> IoC -> Unit Testing -> Development environment (IDE, code tools, code gen) -> Support (documentation, community). I feel like that stack is essential to consistently producing good applications built using the Flex SDK, and I think other companies realize this too.

If you look at what Microsoft is doing with Silverlight, they are positioning themselves to own that stack, and not only that, but they have the advantage of not being the first in that they can fix all the things that Adobe has done wrong. It all looks really promising except for a few things, which I’ll discuss in a moment.

Let me begin by saying that contrary to popular belief, I don’t necessarily enjoy being married to Adobe and the Flash Platform. I really do embrace the ideology of an open, standards-driven web. The market has helped drive Flex because you really could deliver a richer “desktop-like experience” on the web, spawning the RIA buzzword. HTML5 and browser implementations are undeniably slower to progress than a company-owned platform like Adobe has. It was the evolution of Flash as a vector graphics rendering plugin for graphically fun and interesting content that facilitated it’s insane propagation, not it’s ability to draw text inputs and combo boxes.

Ok, back to Silverlight. First bad thing: you are locked into Microsoft. That’s a bad thing, just like being locked into Adobe is a bad thing. However, I would argue that specifically for web development, being locked into Microsoft is better than being locked into Adobe. They own the whole stack, and if you drink the MS Kool-Aid you never have to really leave Visual Studio. Also, Silverlight plays much nicer than Flash with the HTML DOM, which means you as the developer have a greater level of flexibility in how you implement Silverlight as part of your solution.

Second bad thing: you have to work in Windows. It’s not just that I really dislike working in Windows; it’s more that I dislike that I’m forced to. At least Adobe provides Windows, Mac, and (kinda) Linux support for their IDE and Player. I sincerely believe that a web development platform should be OS agnostic.

Ok, wasn’t I writing about Sproutcore? Oh yeah. I’m already way past the tl;dr threshold for most people, but there’s a lot to blab about discuss. All that background was to make the point that an application development platform is more than the display layer and widgets, and even goes beyond the language you’re using. I beleive the fundamentals of an application development platform are:

1. A prescriptive methodology: You or other developers need to be able to look at your source and not have to figure out how you’ve interpreted solutions to basic problems.
2. Consistent development workflow: You or other developers should be able to intrinsically understand how your application can be scaled or altered as it evolves.
3. MVC methodology: You or other developers should be able to intrinsically understand how your application separates concerns.
4. Data access consistency: You or other developers should strive to unify data access methodology so that applications are easier to debug regardless of server technology.
5. Unit tests: The application platform should provide a way of structuring unit tests for all aspects of the platform, including establishing Fixtures for data access logic.
6. Build tools: Scaffolding, configuration, etc. tools should minimize “monkey work” and give the developer the ability to go from thought to implementation as quickly as possible.
7. Good documentation: it goes without saying that docs should ideally help the new learner ramp up, and serve as a reference to experienced devs.
8. Helpful community: I could write an entire post about this one based on my experiences in the Ruby community, but it is critical to have mentors and people who are willing to accept that fact that you are an annoying little n00b and require hand holding like a little child as you ask questions and sound like an idiot for a few days. Also, when you are off the teet and developing like a madman, you need intelligent people to bounce ideas off of. If you go from n00b to the smartest one in the community in a week, you’ve got problems with the community.

Granted, I’ve only been evaluating Sproutcore as I’ve been finding time around regular work, holidays, and spending time with my family, but I’m definitely over the 0 – 1 level of proficiency and actually starting to build things and ask real questions, and I can say that so far, when I need to do something, it feels like Sproutcore has provided for it in a very logical way. All my personal requirements for a good framework are accounted for. I’m not going to port HydraFramework to Sproutcore, because it’s unnecessary in Sproutcore; it already has a wonderful MVC and bindings methodology. I’m not scrounging for a unit testing methodology; it also has been provided for. Skinning / theming couldn’t be easier. Build tools are great. I can develop on my 486 running Ubuntu if I wanted to. There’s always someone idling in #sproutcore who is helpful. The docs are great. It may not have achieved feature parity with the Flex development stack yet but it shows tremendous promise.

So again, the question isn’t Sproutcore vs. jQuery. In fact, you can (and probably will) use jQuery if you develop Sproutcore apps. The question really isn’t even Sproutcore vs. Flash vs. Silverlight, but it’s so easy to frame it up like that. The point of Sproutcore is that it provides the beginnings of an end-to-end application development strategy in HTML5 that competes with Adobe’s Flash Platform (Flash + Flex + Cairngorm + FlexUnit) and Silverlight’s presentation layer and tools.

There are a few additional things I want to point out. I’m not saying it’s the only way by any means. You can write an open standards desktop experience using any recipe you’d like. Maybe my point is that I feel that choosing an application development platform is more than the sum of its parts.

I have also been evaluating Cappuccino by 280 North. For all intents and purposes, it competes directly with Sproutcore, and I’m not blatantly ignoring it. However, my #1 hangup with it is the Objective-J abstraction. I understand using Objective-C for Cocoa development. I understand JavaScript for web development. But to me, I see not much more than a syntactical advantage for Cocoa developers as the point of Objective-J. For me, it makes more sense to stick with JavaScript as JavaScript. I’m going to try to keep my eye on it, especially the Atlas project.

I’d definitely be interested in hearing your experiences with Sproutcore or your own recipe for RIA *choke* or Desktop Experience type applications. I’ll also try to post more details as I continue evaluating.