From my viewpoint it comes down to manageability. I've got scripts that make it take < 2 minutes to set up a new domain or subdomain on my server, and each on

From my viewpoint it comes down to manageability. I've got scripts that make it take < 2 minutes to set up a new domain or subdomain on my server, and each one of those is in it's own directory so it can be copied, backed up, or moved to another server easily. If, on the other hand, you're using somebody else's hosting system, subdomains could be a big pain. For one project of mine, I'm keeping my images on a separate domain so I can switch to a different system for hosting images... For instance, host them on a single-threaded web server or move them to S3/CloudFront. Similarly, if a project involves different technology than the main site or if it may involve different technology in the future, that's a good reason for a subdomain. If you might want to host it differently, that's a good reason for a subdomain. I don't know what to tell you about S.E.O.; a few years ago there were some people who made subdomains by the hundred because they thought they got S.E.O. advantages from that -- S.E.'s like keywords in the domain. My understanding is that keywords in subdomains used to be unreasonably powerful, people exploited them, and now there's not so great. S.E.'s also appear to have depth limits on the crawls they do a single host, so multiple hosts might also be a way to get more pages indexed. On the other hand, there are powerful ways to get monolithic sites indexed too. Other people think it's better to put all your eggs in one basket so you can get a higher score on one host. The truth is that people outside the S.E.'s don't know what the real rules are, and the S.E.'s will change them next year.