Sources:

www: actual content of main site.
blog: actual content of blog.
download: content of download site (only installer pages and such)
*/all.rkt: files that just require all the necessary modules to build the whole site or sub-sites
*/resources.rkt: files that define the resources for a site (icon, css, logo)
minis: smaller one-source-file sites
stubs: template "sites" that are intended to be hooked into other systems to get the racket look.

To build:

Run "./all.rkt" to build all pages.
Run any "*.rkt" to build a page and things that it references. (Use `-h' as usual.)
You can choose -w (the default) for web mode, -l for local using "file://" references, or -r for' local mode using relative references. Normally, you'll want to use -l for testing, and then use -w for deployment.
Use -o <dir> to specify a directory where the built contents is placed, otherwise the content will be placed in the current directory. As a safety measure, the target directory must not overlap with any installed directory. (Use -f in scripts to avoid answering the question about deleting existing files.)
Set the $GIT_DIR environment variable to point to the ".git" directory of a Racket repository if you want to extract release information from a repository other than the enclosing one.

You'll need credentials to upload to S3, and those credentials should be in "~/.aws-keys".
You'll need the "s3-sync" Racket package installed.
Run the "sync.rkt" script: racket -l racket-lang-org/sync