We have chosen to separate the FTP server from the webserver mainly for two reasons (both of which are related to the "separation of concerns" principle):
- The FTP server is mainly meant to be used for data that does not require interaction, and mostly for data that generates a lot of network load. By not serving this information from the webserver we improve the interactivity of other visitors who just want to read small pages.
- A webserver needs to be more reliable, which leads to higher costs of implementation (e.g. redundant disks). Because a lot of the content of the FTP server is mirrored, we can eliminate a lot of this complexity by simply allowing the reliability to suffer a bit.