At the risk of becoming part of something trendy (which my nerd contract expressly prohibits), I’ve finally ditched Wordpress in favor of static solution. Unlike everyone else though I’m not ditching it for performance reasons, because honestly, I don’t get enough traffic to bring down London bridge, which, according to a song I once heard is constantly falling down.
No, what prompted this change was my Wordpress install getting hacked for a third time (to inject spam links, of course). It’s not entirely Wordpress’s fault; each time my version was out of date, despite Wordpress making updating to the latest version super easy. Unfortunately I don’t go the Wordpress admin panel every day to check for updates, so I’m routinely out of date.
All of which to say, I needed something other than Wordpress to get around my update delinquency. Specifically here’s what I wanted:
Little to no code running on the server, with a strong preference for none. No database, either.
It should generate a static site, from flat text files.
It should be able to import all my Wordpress entries
Support the Wordpress site/link structure so that it wouldn’t break existing links into the site.
And a couple of nice to haves:
Support some version of Wordpress themes, or at least make it easy to port them.
Work with MarsEdit
In other words, I wanted Wordpress without the Wordpress.
There were a couple of solutions that got close, but none that did all that I wanted. So, being an engineer, I decided to write my own, which I codenamed Postword. I actually started the project back the last time I got hacked, and even used an early prototype for the blog on Hearts & Tactics.
Postword is more or less a collection of Ruby scripts. It has one to take a Wordpress XML file and convert it into a bunch of flat files (one for each post) that can be processed later by the build script. The flat files are nothing special; they have mail like headers for the categories and title, and then body which is just text with some HTML markup allowed. The build script takes these flat files and builds a full site that looks like Wordpress site structure wise. The script is unique in that it supports the idea of themes, whose API just happens to be the exact subset of Wordpress APIs that my custom Wordpress theme happens to use.
And that’s all Postword was up until a couple of days ago. But if was going to use Postword on this site, which I update somewhat frequently, I needed to reduce the friction. The first thing to do was to write a publish script that pushed the local static site to my server via FTP. I wrote my own script in Ruby, which actually turned out to be not that bad, since I could make a lot of assumptions. The important one was that the local site was always the truth. Which meant if the local and remote ever differed, the local overwrites the remote, and if the remote had a file which the local didn’t the remote gets deleted. In effect, the site becomes self-healing.
The last piece of the puzzle was create a local server that MarsEdit could talk to. MarsEdit supports a lot of different APIs, so I had to decide which one to implement. Wordpress was an obvious choice, but it offered a lot of functionality to I didn’t have or want, plus it was more complex than the others. I decided on MetaWeblog, with the Blogger API as a fall back for when MarsEdit tried to delete a post. It was surprisingly easy to implement, since almost all of the functionality was already in the other scripts, and Ruby provides an XMLRPC server out of the box.
The really nice part of all this is I now have the entire source to this site (flat post files + images + build scripts) in one Mercurial repository. I can download that and publish from any machine, as well as have instant back ups.
Not to say Postword isn’t without its flaws. The build script is slow; it takes a little while to generate all those archive pages that Wordpress creates on the fly. And if the build script is slow the publish script is glacial. When posting from MarsEdit, I can’t actually call the publish function after building the site, because MarsEdit will time out on the HTTP connection. That said, the pros easily outweigh the cons for me.
All in all, I’m tickled pink about how Postword turned out.