Consider adding support for streamed/chunked download of related datasets

[tomachalek]

This would allow for better user experience as even larger datasets would be downloaded without being compromised by multiple involved timeouts for single requests.