I’ve been planing this for a while, but I’ve finally gone ahead and done it. I ditched my previous site, which was built as an amalgamation of Serendipity (S9Y) and Zend Framework (ZF) and moved to WordPress.

Why? Mostly because the previous setup still required quite a bit of work for me to be happy with it: it looked OK for a visitor, but integrating S9Y into an MVC framework like Zend was somewhat nightmarish, and the admin interface was quite cumbersome even for simple things like adding a post. So I’m setting up a vanilla WordPress installation (apologies about the design; I’m working on it!) which will hopefuly translate into more interesting content and less hassle for me!

I initially set up a ZF site to try and get some more experience with the framework itself, and I am very glad I did. It was a completely overkill application of the technology but I feel I can now use ZF for quite more complex applications: it is very capable and powerful while avoiding the typical “application server” setup that seems so common with larger frameworks. I especially liked the way you can cherry-pick which parts of the framework are of interest, including whether or not the whole MVC stack gets used. For simpler applications something like Cake PHP might still be a faster and simpler way to do things, and I’ll keep to plain-no-frills-PHP for really simple things like prototyping.

As for S9Y, it was a bit of a disappointment. I was looking for a blogging engine that could be plugged into my ZF setup, but since there seem to be none that are happy to do that kind of “blog backend” role only, I settled for something that seemed easy to hack into submission. S9Y does bring its own kitchen sink as well, but it seemed easy to hack. However, two things let it down: convoluted, poorly-documented code, and the templating system.

I won’t say much about the code quality, except that it was at first sight simple but in fact quite messy once you tried to wok out interactions between different parts. The lack of documentation was the really frustrating part, though. So far WordPress seems much better documented and understandable.

The templating system of S9Y is Smarty. There’s nothing wrong with Smarty itself; I guess my problem is with templating systems in general. I know these have their usefulness in some development environments, and of course it is essential to separate as much logic as possible from your markup. However, I don’t think templating systems are a great solution: many times the behaviour of a template is itself complex, and a custom-syntax, crippled language like Smarty is a hindrance.

So there you have the reasons for moving on. Both ZF and S9Y had its fun moments but that setup wasn’t really suitable for this tiny site. Getting rid of the not-invented-here syndrome is also a good thing. In the coming days/weeks I hope to put together a nicer design (I promise I’ll go easy on the orange…) and I now also have fewer excuses to create content more often. So overall I think this was the right decision!

Despite Adobe and Microsoft probably wishing otherwise, the vast majority of the web is still built using trusty ol’ HTML, with its now inseparable partners CSS and Javascript. The latest valid specification of HTML (4.01) will celebrate a decade of existence by the end of this year, so it seems appropriate that a shiny new version is in the works.

Since the web is the vehicle for the ongoing information revolution (which itself is transforming so many areas of the lives of billions of people all over the globe), it is important that the basic building block of the web – HTML – is up to the task: it must continue to support rapid, exciting, and unpredictable developments in the web. It’s a lot of responsibility. Enter HTML 5.

There is a lot on the current Working Draft, so instead of listing all new features and changes (plenty of other sites do that already) I thought I’d share my thoughts about some of the features instead.

Good stuff in the current Working Draft

The lexer and parsing rules. These are great, even if their true usefulness for “serious” web development might be somewhat limited. Now any syntax error on the HTML code will result in the same final DOM and hence the same base model for rendering the page – so HTML rendering will be consistent (or at least consistently broken) across browsers.

Focus on the DOM. This is closely related with the above point. HTML 5 defines both the syntax of the new language but also the structure of the DOM used internally by the browser (and the rules that DOM must follow). In theory, this should bring about much more cross-browser coherence.

New APIs (drag & drop, offline and storage support, canvas, media control). Each one of these is worthy of it’s own blog post, but here’s the shortest summary possible:

• If you ever tried handling drag & drop on a browser you would probably have offered one of your own lungs to try and make the process easier. Even the existing Javascript library solutions suffer from quite bad performance for this most common of UI operations. Having in-browser support is a sure winner.
• Offline state and storage support are essential for reliable web applications and can truly bring about the slow death of fat desktop environments (mmmn, I wonder when will Microsoft implement this in IE…)
• Canvas and media control APIs are essentially the second step towards Flash/Silverlight without Flash/Silverlight: complex UI widgets built with Javascript. (The first step, by the way, was simple animations provided by JS libraries.)

Removal of presentational elements and attributes. Good riddance. This might actually focus some minds into separating their markup from presentation. Now, if only there was a way to get rid of tables-for-layout as a language feature…

Web sockets. Goodbye polling, hello highly-specialised, highly-efficient web applications. This will put current Comet techniques to shame.

Not quite-so-good stuff

Lots of new arbitrary elements and attributes. This is my pet peeve with HTML 5. What is a sensible new element today might be redundant tomorrow as the patterns of semantic organisation change. Furthermore, there are an incredibly large number of other potential elements that would make much more sense to specific areas of work but aren’t included. So we will get <head> and <article> and <aside> and <datagrid> and another thirty-something (at the last count) new elements and attributes. I believe this is the wrong solution, where we are forcing very specific meanings into the language that will be both useless in a large variety of cases and deprecated in the not-so-far future.

The main issue that developers seem to battle with is how to apply a specific semantic model to the language. By adding new elements with pre-determined semantic meanings we are perhaps solving some of the common problems but have not addressed the root causes of the problem: we need custom semantic models. I believe XML has solved this problem with DTDs and namespacing, but those solutions fail for the web due to complexity.

I am not offering a solution to the above problem, I am simply pointing out that the new tags are the quick solution, not the best one. Maybe we could use something akin to annotations in Java? Anyway, that’s a topic for a later blog post…

Editing API. It’s a good idea, but the current specification amounts to little more than a roughly outlined wishlist of a concept. This is not to bash (hard) work-in-progress, but I don’t think the HTML 5 language itself should be concerned with defining the behaviour of a GUI IDE for itself. This is a big project in its own right.

It’s too big. The current Working Draft looks too much like a huge wishlist, with many features being defined within the spec itself. Smaller specs should be split into their own documents and HTML 5 should simply specify its dependencies on those documents. This includes the rather verbose “common microsyntaxes”, several parsing rules (URLs, content-sniffing), canvas API, etc. Trimming things down should also simplify adoption and allow for a more solid spec. For instance, if a given dependency is not defined solidly enough it should be easier to depend on a simpler/earlier version, or even scrap the dependency altogether in favour of having a well-defined HTML 5 spec.

Web sockets. Yes, I listed this above in the “good stuff” section. It also has the potential to put too much power in the hands of web developers (not always famed with good application design). Seriously, though, the web has flourished on top of the HTTP protocol, and although this feature will probably be a good thing it must be treated cautiously due to introducing a different paradigm for data transmission over the web.

Why isn’t there more buzz?

There has been a fair amount of discussion regarding HTML 5, but given the fundamental aspect of what is being changed I guess I expected some more enthusiasm (or hype). Partially I believe this is not happening because of the scale of the project and the fact that it’s still a far-off prospect. “Cool” (OK, let’s call them “popular”) new web technologies seem to generate buzz and grow in adoption in a matter of a few months. They tend to be something that developers can simply pick up and use: microformats, embedded maps, web service APIs, Javascript libraries… HTML 5 is not something that you can use right now and reap any significant benefits from.

The other aspect is that the current HTML implementation is not such a big problem at the moment. Most of the quirks are well-honed, and solutions to problems have long been known and worked around. Most of the bigger problem faced by web developers are related to either CSS or Javascript. The truth is, the current flavour of HTML has withstood the last ten years well because it does it task remarkably well: it provides a stable infrastructure for defining the semantics of content, over which the presentational (CSS) and behaviour (Javascript) layers can work.

In any case, this is definitely a worthwhile project which will eventually offer some real benefits to the world of web development even if by itself it will probably not be the driving factor of a great information revolution – that has been with us for almost ten years now.