On 3/6/07, Brody Pierre <[EMAIL PROTECTED]> wrote: > So a blog-server admin would naturally want to host many bloggers. I'm > thinking of something like blogger.com , although presumably specialized for > some niche. > > What motivated the decision to support one blogs (now moving forward to > supporting a few)? What are the use cases that made Typo and other blogging > engines work this way?
I'm guessing that it's because building thread-safe code which can deal with hundreds of simultaneous requests for hundreds of different users' blogs, and do it all securely, is hard. Much easier to set up a separate Unix account per customer. That way you get: - Stronger security, because it's enforced at OS level. You don't need to worry (as much) about user X injecting some SQL and wiping out 200 customers' blogs from your database. - Better ways to manage disk and CPU usage (e.g. run 'and' to stop one user's blog soaking up all CPU time) - Ability for users to add their own CGI, deploy their own applications - Ability for users to FTP/FTPS in and edit their web pages and templates directly (*) - Ability to scale indefinitely, as it's relatively easy to move users to less loaded servers transparently if you have to Building all that into the application layer is a lot of work requiring lots of attention to detail. People like Google and SixApart have multiple full-time developers dedicated to their blog software, dealing with security issues that crop up, load issues, and so on. The (*) shouldn't be downplayed either. Users expect it. Even cheap $10 hosting from GoDaddy comes with a Unix account and FTP access, so why not use those accounts to segment your hosted blogs too? Basically, unless blogs are your entire business, I think you'd be crazy to invest in the amount of work required to support lots of users on a single software deployment. mathew _______________________________________________ Typo-list mailing list [email protected] http://rubyforge.org/mailman/listinfo/typo-list
