git commit

September 23, 2010

Around the beginning of July, my hosting package with A Small Orange was expiring. The company had treated me well, but I was becoming unhappy with hosting on a shared server. I had completed a project earlier in the year that involved hosting Django on a Linode instance. This whet my appetite for more Django work, but I didn’t feel that it made sense to host a tiny, never-updated blog on a VPS I would later have to babysit. Instead, I looked to Google App Engine.

I’d worked with App Engine on a couple of small personal projects – toys, really – and one medium-sized site for work. I liked using the provided web framework, but hadn’t been successful at getting Django really working on it. This time, though… well, at least I got it working. There were all sorts of fiddly issues – no PIL for image manipulation (even though I needed PIL on my local machine for emulating the production environment), gaps in the Django to Google Accounts authentication bridge, and a number of other small nit-picks. I decided that I could just go forward with the Google provided framework, but at this point I was also a bit disheartened. I wanted to look around for other options.

After a time, and many reviews read, I found Webfaction. I was told by all the reviews I encountered that they were the best Django shared host around. I gave them a shot, but they didn’t fit what I was looking for. Everything about the service was extremely flexible, but there was a strange disjointedness. There were many things which could be done through the administration panel. For the most part, admin panel tasks were straight-forward and Just Worked(tm). However, there were often tasks which required that you do a thing in the admin panel, SSH into your account to twiddle some scripts, then do another thing in the admin panel to finish the task. I felt like, if I already had to SSH in to get things done, why not just pay slightly more and have a full-on VPS? Of course – again – I didn’t really want a VPS just for a silly blog.

Eventually I cancelled my account and was given a full refund. I really don’t want to seem too harsh on Webfaction, as I didn’t really give them a fair shake. I just showed up, saw some things which rubbed me the wrong way, and left. By the amount of others who can speak no ill of them, I’m sure they’re doing something right. However, I was realising at this point that I just wanted something simple and very light. I also knew that I wanted to have absolute control of my markup, so a Wordpress or TypePad sort of thing wasn’t going to cut it. By this point, I was feeling sort of fed up with the whole personal site thing, and I stopped looking around for a while.

It wasn’t until quite recently that I stumbled across toto. This reminded me of Osxome, which I was using on A Small Orange to host the previous version of the site. Now, Osxome isn’t very good, and I can say this with some authority, being the one who wrote it. It worked passably for its primary use, getting the North Knight Software site up quickly and cheaply, but for everything else it was too cludgy and unpolished. toto on the other hand looked pretty slick. It was developed for use on an amazing looking service – namely Heroku, it uses Markdown instead of plain HTML, and (most importantly) it’s designed to use git to move posts to the server. I wish to heck I had’ve thought of this.

So, yes, I could have dusted off Osxome and tried to make it not suck, but I decided instead to give toto and Heroku a shot. That’s where I am right now as of this writing. Hopefully I’ll not lose the blogging urge for a little while yet!