I was in a fascinating seminar by Barney Boisvert about programming with "progressive enhancement". It's something I've thought about on the edges of my mind, but Barney (who is funny and really tall) put it so clearly that I could not help but blog about it. To get us in the right frame of mind, let's check in on project launch meeting at CF Webtools. Our new client, Joe Dreamer, has brought his latest idea to the table. He wants to build a special "Barbie's Facebook" - a social networking site where young girls can register their Barbie dolls and flesh out their lives in the virtual world. NOTE: This funny example came out of the Muse' head. I know of no such project (but now that I think about it....). Anyway, let's check in on our project launch meeting.
Meeting attended by the Muse, Laurie (our gifted project manager), and 2 developers - Pinky and The Brain.
Buried in our little sketch is the outline of one of the typical processes we go through when designing web applications. Note: I'm not talking about agile, scrum, waterfall or any other project management idea here. I'm talking about our choices when we actually build the application. Now I know many of my smart and savvy readers will say "browser compatibility should be a part of the spec." And indeed compatibility and testing usually are part of the spec. But our actual coding choices tend to operate in a way that is not productive. We typically build our sites as a pyramid but from the top down. This is where Barney really helped clarify it for me - using small words so I could understand. He defined the pyramid and suggested that we start with the base and work our way up to the ideal. Before we get there however let's discuss his phrase for the way we often do things - graceful degradation.
Here's what oftenhappens. We start with a storyboard, a workflow and some design concepts. These are important because they are what the stakeholders are concerned about. But it's important to note that while story boards, mockups and design concepts are foundational to our project planning, they are in fact representative of the ideal that we are trying to accomplish. In other words they picture the top of the pyramid. But they fix in our minds a vision of how the end product should look and behave. Since we are developers we have the latest Firefox, Safari and Chrome browsers - and most of us open IE only to test (and occasionally de-test). Our goal is to build our application to the highest possible standard - to build that ideal we see at the top of the pyramid. That means nice Web 2.0 interactions, client side error traps, rich images, transparency, dialogue boxes, jQuery grids and panes etc etc - the works. In a single phrase our approach is to "build the ideal first".
Often our approach (the "graceful degradation" approach) is to build the ideal and then handle the exceptions as they arise. So we build a fancy pants interface, test, congratulate ourselves and then after the fact we have to account for Aunt Martha's custom Gateway IE 4.0. It's either that or teach her how to use an new operating system and browser and we are too conscientious to do that to the local suicide rate. The result? Our code ends up with hacks and work arounds and widgets all designed to solve specific issues we encounter after we build the ideal.
Note, we are talking about web applications here. Rich Internet applications (RIA) using something like Flex or Silverlight are designed for a single runtime that is required - so there's no degradation at all. It either works or it doesn't.
Recently my wife and I went to an upscale steakhouse (Mahogany's) for a meal. You know you are in a luxury restaurant when the hostess looks carefully at your wife and then substitutes a black napkin for a white one to better match her outfit. It seems like there were about 3 wait staff who hovered a few feet away and the minute I dropped a crumb on the table they were there to sweep it up with the little crumb blade thingy. This restaurant had a mostly a la carte type menu. You choose your meat first (that's why you are there after all). Then you add a side or two and a bottle of wine and possibly desert. The focus of the meal is the meat. The waiter spends 2 minutes just describing the process of selecting and preparing your steak - "First we raise a cow lovingly with chocolates and bubble bath, then we slaughter it carefully with a 200 year old Katana, then we yada yada yada and....", you get the idea. It's clear they are quite proud of their meat - which is after all an American tradition (at least among men). The side dishes and salad were great too. In fact I had a "salt encrusted" baked potato that was absolutely delectable. But the sides only enhance the main event - the main course. In this place steak rules and it's very very good. If all you ordered was the steak and a glass of water you would have a memorable meal - but the sides, salad, appetizers and wine take it from being a meal to a grand evening event.
If you are following my metaphor, I mean to say that using progressive enhancement there is a basic utility that should be considered and created first in each web application. Think of it as the "main course" - the basic functionality you are trying to deliver without the bells and whistles. This baseline should be able to provide the functionality of the application to the lowest common denominator device or browser you intend to support. How do you get to the "lowest common denominator"? Here's where Barney had some specific advice for us:
Once you have your main course it's time to start adding the sides. Semantic Markup, HTML 5, fancy pants CSS, and of course the rich tapestry of jQuery (or one of the other JS libraries) that allow for page injection, DOM manipulation, and functional targeting. Continue assembling the pieces until your meal has turned from a simple main course into an evening event that reflects your "ideal" - that thing you sold the stakeholders on. But the good news is you will have far less to do now for browser compatibility and you won't be caught with your pants down on a browser that doesn't work. They will all work (although they will not all reflect the ideal).
Overall I like this approach. There are times when building for graceful degradation (building the ideal first) is still ok, like an application for a well-defined audience. If you know all your users are internal and have a certain browser, or can be required to use a certain browser, then you can program to the capabilities of that browser. But for public sites that are designed to cast the widest possible net and have a wide demographic (or a demographic of "late adapters") the progressive enhancement approach will mean better regression and less headaches overall.
I'll leave you with this quote from Barney's presentation (which can be found at this link).