Tuesday, November 15, 2011

simplicity simplicity simplicity -- shouldn't that just be 'simplicity'?

The other day I was thinking of Richard Gabriel's classic The Rise of 'Worse is Better'. The article has a kind of old context, but I think its core is still valid. (I also appreciate the vacillation within the article; these are issues about style and mood where there won't be a single correct answer.)

As the Wikipedia Page reiterates, the core components of "Worse is Better" are, in roughly descending order:
The design must be simple, both in implementation and interface. It is more important for the implementation to be simple than the interface. Simplicity is the most important consideration in a design.
The design must be correct in all observable aspects. It is slightly better to be simple than correct.
The design must not be overly inconsistent. Consistency can be sacrificed for simplicity in some cases, but it is better to drop those parts of the design that deal with less common circumstances than to introduce either complexity or inconsistency in the implementation.
The design must cover as many important situations as is practical. All reasonably expected cases should be covered. Completeness can be sacrificed in favor of any other quality. In fact, completeness must be sacrificed whenever implementation simplicity is jeopardized. Consistency can be sacrificed to achieve completeness if simplicity is retained; especially worthless is consistency of interface.
The essay goes on to explain why these factors can be crucial to a technology really taking off. (Also it talks about the "New Jersey approach" vs the "MIT approach", terms I am including here for my future grepping needs.)

I find a rough analogy in my professional life as a UI Engineer when it comes to the selection of toolkits. I had a miserable time with "Wicket" and its bizarro blend of delicate flower Java class/HTML pairs. (It was barely ok for light work, but god help you in having to learn this whole lifecycle flow if something went wrong or you wanted to make your own component.) These days I've been arguing sticking with basic jQuery rather than taking on some heavier duty toolset like YUI or jsMVC.

So I ask myself, am I just a stick-in-the-mud curmudgeon who hates to learn new stuff? That's at least partially true: I'm much more interested in what technology can empower me to build than in the tools themselves, so I'm only interested in new toolsets when they let me do something new or in a vastly easier way, not just something old in a slightly more concise way. But I think there's something deeper in my preference for light weight toolkits, more profound...

I've said "People and computers should be judged by what they do, not by what (you think) they are." No one cares how finely crafted the objects in your code are, they just want to know, does it do the job well, and have you set yourself up to do the next few, slightly different jobs well. When I'm debugging, it's because the system is DOING something wrong, and I need to correct that behavior. My tools are-- lo and behold-- a debugger! and thoughtful logging... (good old log-based debugging is a crucial tool. I say that's because it forces you to think about what your preconceptions are at each step of the way, and then show them in action. But there's a chance it might also be because the debugging tools I used in college were so awful and opaque. Also, debuggers mess with the timing of stuff.)

Anyway, when I'm debugging, in a debugger, if I'm mostly looking at other's people code in the stacktrace, my job is much, much tougher. Most toolkits are extremely configuration heavy. (Toolkits by their nature don't just solve YOUR problem, but a hundred other problems as well, so you have to specify lots of things) It takes a fair amount of expertise and learning to set them up right, which costs time-- time you often don't have, the whole reason you're using the toolkit is for it to quickly solve your issue-- and when things break, there's often a huge gap between the misbehaving code and your error. The lighter the toolkit, the more likely the mistake is near where the debugger is showing the problem, the nearer the observed effect is from the code that caused it.

JQuery, I hardly ever have super tough to understand problems with. The few times the issue shows up in "jquery.min.js" itself, its been where I'm setting some property IE thinks should be read only or some such. Other tools I used, like Java RichFaces, the situation was much, much hairier. But with basic Javascript calling jQuery when needed, I can get to the problematic code easily.

Now this argument could be taken to an illogical extreme. I'm not saying you should code in assembly or anything like that... but your toolkits should be simple, reliable, and your widgets and libraries should be of the do one (hopefully difficult) thing and do it well.

(Of course, someone pointed out another reason to hate toolkits is they put a coating over the 90% of coding that is easy and pleasant and 'hey I built this!' and leave behind the 10% that is difficult anyway, and how you earn your pay... in fact they usually make that 10% significantly harder.)

Continuing the comparison to assembly, you still get people trying to spin Javascript as the Assembly language of the web, the code that all browsers speak, and it shows up in your debugger, but not what you want to be coding in. This kind of thinking is weak, because the additional layers, while slick and clever (epecially in the early baby tutorials) don't actually let you say stuff new. A concise essay about that in terms of toolkits is Uriel Katz' Why JavaScript is NOT the new Assembly  and a longer one against CoffeeScript and other pre-processor type languages is blog.izs.me's "JavaScript is Not Web Assembly".

No comments:

Post a Comment