Re: Accessing yum-archive.postgresql.org

That makes sense. S3 isn't exactly helpful about these things. 

For interactive use, yes. But if the infrastructure has to generate files, then easy scriptability would be really good too. HTML isn't the greatest for that unless you mix in XML with a secondary namespace, and then you start upsettting browsers.

You also probably don't want to be re-writing one giant index.html or Packages.xz or similar manifest file each time the archive is updated. 

So I think side-by-side HTML and some kind of json or similar would make sense, both generated and updated at the same time in the same job. There'd probably want to be a master listing entry point plus sub-listings, both for the HTML and script-friendly representations. I don't really have a strong opinion about the format of such listings so long as there's a browser-navigible HTML variant and something easily scripted.  I'd lean toward json since it's so widespread and convenient, but pretty much anything right down to a tab separated flat file would do.