On Jan 25, 2008, at 9:34 AM, Saimon Moore wrote: > Steps on 3rd request: http://es.example.com > > 1. WS: can't find index.html in /var/www/apps/example.com/public/es <= > Note the doc root is different > 2. WS forwards request to mongrel process (M) > 3. M finds index.html in /var/www/apps/example.com/public > 4. Mongrel serves /var/www/apps/example.com/public/index.html which is > the page associated with the http://example.com domain. > 5. This is NOT the required behaviour as it's serving the cached file > for http://example.com rather than http://es.example.com as was > requested.
Wait... my brain is still missing something; I apologize. Why isn't nginx sending the stuff from es.example.com to mongrel.example.com/es? You don't *need* to send your proxied requests to mongrel's root... do you? I guess you'd need to override the behavior of Rails's url_for() or generate routes asymmetrically, which would suck. My gut (as a non-devteam-member) is that you're doing something a little strange, so a custom handler is more sane than patching a switch onto Mongrel. A more interesting proposal would be to say "Mongrel is ONLY an app server and doesn't serve static files AT ALL" -- to make it do so would require a custom handler. That, however, would be a Big Scary Change (TM) and I think users would be... unhappy. -Nate _______________________________________________ Mongrel-users mailing list [email protected] http://rubyforge.org/mailman/listinfo/mongrel-users
