Serving

Important

Make sure to read Requirements before proceeding here.

Using S3

Once binaries are pushed to S3 and package index files are created, you can use them with install.packages() and alikes. When setting the repository, provide the URL to the directory containing the package index files (e.g., PACKAGES):

install.packages(<pkg>, repo = "https://<s3 url>/")
# Example
# install.packages("brew", repo = "https://cran.rpkgs.com/arm64/noble/latest")

Tip

While this works with the default S3 URL, you'll typically want a cleaner URL than S3's default. Create a CNAME DNS entry from your domain pointing to the S3 directory URL containing the package index files.

Using a Content Delivery Network (CDN)

Installing packages directly from S3 has some downsides:

• Files are downloaded fully for every request
• Download times increase with distance from the S3 region (and usually packages are only stored in one region)

A CDN in front of the S3 bucket solves these issues. It comes with following benefits:

• CDNs cache files from the underlying S3 source after the first request for a specific period. Subsequent requests for the same file are served from the CDN cache without querying S3. This is faster and reduces S3 request costs.
• CDNs distribute files across multiple edge nodes globally instead of serving them from a single location. Users will receive the assets from the nearest regional server and its respective local cache, reducing latency regardless of location.

Note

CDNs are specialized in delivering content from a central source as quickly as possible worldwide. Whether your packages target global users or just yourself, a CDN will always serve binaries faster.

Tip

CDN usage incurs costs that vary by provider. However, CDNs do not always increase costs: many offer generous free tiers, and by reducing S3 requests, they might actually cost less than when serving assets directly from S3.

(Just make sure to not use the most expensive CDN out there 😉️)