lighttpd.net runs on 1.5.0

All the lighttpd.net domains (blog, trac, www, xcache, upload, …) are now running lighttpd 1.5.0-trunk. It took some debugging to sort out problems with the way trac wants to handle PATH_INFO r1841 and how ruby handles fastcgi-keepalive requests via Unix-Sockets r1850 + r1849.

In case you see that one of the sites is down, ping me (weigon) on IRC. The server is running in valgrind and should provide enough information to fix the problem.

compression

All the content is sent as compressed content if the client supports it. As some browsers have problems with compressed javascript and CSS we only enabled it for Firefox for now.

deflate.mimetypes = ( "text/html" )

$HTTP["useragent"] =~ "Firefox" {
    deflate.mimetypes += ( "text/javascript", "text/css" )
}

trac

For the trac I use:

$HTTP["host"] =~ "^trac\.lighttpd\.net" {
    # alias, ...

    $HTTP["url"] =~ "^/trac/" {
        proxy-core.backends = ( "127.0.0.1:9090" )
        proxy-core.protocol = "fastcgi"
        proxy-core.rewrite-request = (
            "_pathinfo" => ( "^/trac(/.*)" => "$1" ),
            "_scriptname" => ( "^(/trac/)" => "$1" )
        )
    }
}

the rewrites tune the PATH_INFO and SCRIPT_NAME settings to the needs of trac. It should also handle the old bug around trac without a prefix.

rails + fastcgi

Thanks to the rails apps we run on blog.lighttpd.net and www.lighttpd.net I could trace down some nasty bugs. One is still open and is around the fastcgi keepalive + fcgi.rb. It isn’t tracked down yet, not enabling it seems to be safe for now.

For mephisto we are using the cleanurl.lua to get rid of the server.error-handler-404.