There’s something about waking up at 3am every morning that just gives you that extra bit of time to blog. Today we’re touching on what happened to my GeoCities Archive. A while back I had an idea to dockerise my Apache-based web server.

There wasn’t much point in having a dedicated piece of hardware when I could just use Docker Desktop on one of my other computers. I began to put pen to paper in early September. The plan was to merge in a better version of the GeoCities Archive and use a case-sensitive APFS volume on macOS.

I rsynced my data with a little help from Jon and it was all looking good. Until I realised how much smut and gore was in the archive. In the material I kept stumbling across, a lot of it appeared to date from after Yahoo!’s acquisition of GeoCities in May 1999. That is an observation from what I was finding in this copy of the archive, not a claim about GeoCities as a whole.

This posed a very big problem for me. While Cloudflare’s CSAM implementation helped with one very specific category, it could not solve the broader moderation problem. The tool compared cached images against hashes of known CSAM. It was not a general pornography, gore or GeoCities ToS scanner, so everything else still needed some other way of being found.

Sure, I could manually search for pornography and gore by filename, but I reckoned I was only finding about 5-10% of it that way. That was my rough impression, not a measured detection rate. Plus I had to look at these sites to assess whether they were worth keeping.

In most cases, they were not worth keeping and I made the decision to pull the plug. Only a few days ago, I reverted back to the Raspberry Pi setup without the GeoCities Archive. I also trimmed away Sci-Fi and Assembler Games. If I remove GeoCities, I can also remove the blade hard drive from the web server.

The good news? Material I found that appeared to violate the original GeoCities ToS was reported to archive.org for review and possible removal. I just hope they do actually purge all the questionable content I had to sift through. It was nice knowing you GeoCities, but perhaps you are best left in the past - in our memories. 🥲

Sources