I’ve been writing Some TeX Developments for ten years now, starting off on WordPress.com before moving to a self-hosted WordPress set up. All of this time, I’ve stuck with WordPress as it’s a very powerful and flexible system. However, it’s got some downsides too. In particular, as it is dynamic, database-driven, system, the pages are created each time someone requests them. That’s great for things like supporting comments, but it means there’s a non-trivial amount of work done each time someone views a page. That turns into a real cost when you are paying for your own hosting. My most recent hosts were really good for support, but I needed enough CPU cycles to push me into the ‘non-trivial’ cost bracket. At the same time, a dynamic site means that there’s always a security risk.
Enter GitHub Pages
I’m hardly the only person to come across these issues, and it’s no surprise that there are a variety of good solutions. One that’s really gained in popularity over recent years is GitHub Pages. This uses a specially-named Git repository to run a generation system called Jekyll. Unlike WordPress, Jekyll generates pages when the sources are committed, so the pages themselves are static ‘classical’ HTML.
Rinse and repeat
To go from WordPress to Jekyll, I started by extracting all of the content using the WordPress to Jekyll Exporter plugin. That gave me a set of HTML files which nearly worked straight away (but with no styling). After a few bits of clean-up to make things work at all, I then did a load of search-and-replace steps. Most of these were to convert the content to Markdown, clean up minor mark-up issues, etc.. I also took the opportunity to work on fixing typos, broken links and so on: that is a lot easier to do with a local set of files, compared to WordPress.
Most of that work was very mechanical, but it took a while: most of that was because of flaws in my original text, not the exporter!
Exporting the content doesn’t deal with the website style, nor does it include comments. The latter don’t work in Jekyll directly, though one can use Disqus. I decided against that for the present: I don’t really need a discussion system for my blog.
Getting the style right could have been sorted by copy-pasting the raw HTML from the old site. But I decided to take the opportunity to revise the layout. At the present, it’s based on the LaTeX Project one, but rather simplified. I may well look at this again, fixing minor issues as I go. But I’m no design expert: I’d be very happy to have suggestions!
The final thing to do is to get the web address sorted. I’m just sorting out with my registrar and GitHub, and that will be done shortly. Hopefully, with that done, my latest blog rearrangements will be done!