> Most of these sites are not even that big. I expect maybe a few thousand visitors per month. Incidentally, if you can make a site "static", so far I'm mostly

> Most of these sites are not even that big. I expect maybe a few thousand visitors per month. Incidentally, if you can make a site "static", so far I'm mostly liking AWS CloudFront loaded from S3. After many years serving my site from a series of VPSs/hosters/colo/bedroom. It's fast and inexpensive, and so far perfectly solid. Deploying consists of updating S3, and then triggering a CloudFront invalidation, which takes several seconds. The two key fragments of my deploy script (not including error checking, etc.), after the Web site generator has spat all the files into a staging directory on my laptop where I can test them as `file:` URLs, are: aws s3 sync \ --profile "$AwsProfile" \ --exclude "*~" \ --delete \ "$WebStagingDir" \ "s3://${S3Bucket}/" and then: aws cloudfront create-invalidation \ --profile "$AwsProfile" \ --distribution-id "$CloudFrontDistId" \ --paths "/*" \ < /dev/null 2>&1 | cat The main thing I don't like about it (other than the initial setup wizards having a couple bugs) is that it doesn't automatically map `foo/` URLs to `foo/index.html` S3 objects. The recommended solution was to use AWS Lambda, which I did temporarily, and it works. But when I get a chance, I will see whether I can make my deploy script duplicate S3 `foo/index.html` as S3 `foo/` and/or `foo`, so that I can get rid of the worse kludge of using Lambda. Unless CloudFront offers a feature to do this before then.