@@ -0,0 +17,4 @@
+
+This was great, set and forget, except for a pesky VM that needed maintenance, patching, and more. So every so often, the downloads site needed to go down so things could get patched. A second VM could have been used to load balance traffic, or any number of possible other solutions could have been chosen, but that would've just added to the overall maintenance burden and cost. So instead of dealing with servers, we went "serverless" (yes, there are still servers, but someone else is dealing with them). The caddy plugin to generate file indexes was replaced with a "serverless"-function that queries the S3 Bucket directory and a CDN that has direct integration with the S3 Bucket so it can serve the files also directly but will fallback to the file listing function if a file isn't found for that route. The CDN has aggressive caching, so the file indexes only need to be generated once. Also, when files are uploaded to the S3 bucket, another function listens for file upload events and clears the cache on a targeted basis to not completely clear the cache.
+
+The serverless functions and infrastructure overview have been open-sourced in case any other project wants to poke at what we've done and use it for themselves. Please feel free to hop into our chat and ask any questions.