I'm writing (and uploading) this about 4 years after the talk don't exactly remember how it all works, but I'll do my best to explain the main points.
This is the code behind my talk at JSConfEU 2017, "What's new in Netscape Navigator 2.0".
The talk was presented using Netscape 2.0 (and 4.0) running in a Windows 98 VM in Virtual Box. I won't cover how to set one of those up here.
The slides were generated from Markdown for convenience - the main source is in slides-src/slides.md. The other files in slides-src are Handlebars templates + supporting JS files for each template type that were used to generate the static slides. The look and feel was based on the Atlassian presentation template at the time - which was probably only amusing to other Atlassian folks. I can't remember how all this hangs together to be honest, lib/slide-generator.js does all the work of taking the markdown and the templates and combining that all into Netscape compatible HTML.
There was a whole thing in there for syntax highlighting code samples - but then having to stick in old school <font> etc tags in order to render correctly in Netscape. I definitely don't remember how all that worked. Looking at the code I generated HTML and then used Cheerio + a CSS parser to get the right colours for each style that could be replaced with <font> tags.. that's all in lib/highlight.js.
One key point: in original JS there were no keyboard events, so to advance slides I needed to click on the "next" link. In addition, I couldn't have speaker notes or anything else - so there was just a one work description of what the next slide was above the link. Thanks to <frameset> I didn't have to move the pointer between clicks to advance the slides. This combined with a lot of practice led to the talk going (hopefully!) smoothly.
The slides themselves were delivered from the host machine via a local webserver. It looks like there was a way to generate static HTML version in genstatic.js - but I can't remember if I just had that as a fallback, or I used it for the talk itself.
The TodoMVC implementation was hand-coded in /public/todo.
There's some other handcrafted HTML/JS in /public that was brought in with the iframe template type.
I think that's all the main bits.. I've probably missed lots of little things that took me ages to work out at the time.
There's some mirrored documentation I mirrored in case the originals went away in archives/. Apologies to whichever University it was that I accidentally DDoS'd mirroring one of those.