httpd documentation 7/16/96 by Michael Temari updated 2006-06-01 by Al Woodhull DISCLAIMER: Use at own risk etc... COMMENTS: Please send me any bug reports, comments, questions, etc... My email address is Michael@TemWare.Com BACKGROUND: httpd is a World Wide Web (WWW) server. I wrote it from scratch so the setup and configuration will not be like other web servers though hopefully by reading this document there will be no problems in getting my web server up and running on your Minix system. COMPILING: To compile httpd all you need to do is type "make" in the httpd source directory. There should be no errors or warnings. If you should run out of memory when compiling try adding the -m option to the CFLAGS list in the Makefile. INSTALLING: To install httpd all you need to do is type "make install" in the httpd source directory. By default the place to install httpd is into /usr/local/bin. If you would like to change this then change BINDIR in the Makefile. Httpd will be linked to in.httpd, which is the preferred name for a program started by the tcpd internet access control program. The program dir2html is also installed -- this provides a directory listing when a web client accesses a directory which does not contain a file named index.html (or an alternative designated in /etc/httpd.conf). The man pages are installed by typing "make installman". CONFIGURING: Before running httpd it must be configured. The name of the default configuration file is /etc/httpd.conf or you may pass the configuration file name to httpd. Upon starting up, httpd will parse the configuration file and then process requests. This README file and the sample httpd.conf may also help in configuring. The httpd.conf.5 man page presents the same information for reference use. The configuration file is an ascii file which consists of lines of the following form: directive LWS [parameters separated by LWS] NOTE: LWS denotes Linear White Space which is spaces and/or tabs The following are valid configuration file directives: serverroot redirect user chroot logfile dbgfile dirsend direxec vhost auth proxyauth vpath include mtype To make the file more readable, on directives which occupy multiple lines you may omit the directive on lines after the first and begin these lines with LWS. serverroot path The serverroot directive sets the translation for // to the given path. redirect url If redirect is defined in the configuration file then all request urls will be redirected. For example, if in the configuration file of minix1.hampshire.edu this line appears: redirect http://minix1.woodhull.com/ a request for http://minix1.hampshire.edu/some/page will return a 301 error which is a redirect permanent to http://minix1.woodhull.com/some/page. user username The user directive causes the server to run as the given username, otherwise the server will run as whoever started it (normally root). chroot directory The chroot directive causes the server to chroot to the given directory after the configuration and log files have been opened. Normally this will be the home directory of the given username in the user directive. NOTE: /~user will be translated to the home directory of the given user // will be translated to the serverroot directory NOTE: if this directive is used then beware of the consequences. logfile filename The logfile directive tells the server where to log http transactions. NOTE: the file must exist to enable logging dbgfile filename The dbgfile directive tells the server where to log debug http transactions. NOTE: the file must exist to enable logging dirsend filelist The dirsend directive tells the server that when a directory is requested that it should send the first file that it finds in the directory from the filelist for the request. direxec program The direxec directive tells the server that when a directory is requested and no file is found from the dirsend directive that it should run the given program. NOTE: the program normally generates a directory listing on the fly NOTE: the program access is considered X with no access restrictions. vhost hostname VhostRoot vhost is for defining access for virtual hosts. If none are configured then any host is accepted. If specified then access is only granted for requests for hosts which are configured here. In the Vpath section below the /// gets translated to the corresponding VhostRoot. auth authname authdescription access [passwdfile [users]] The auth directive sets up different authorizations with the server. The authname is the name given to the authorization and is case insensitive. The authdescription is the description of the authorization and is what the user will see when asked to enter a username and password. The access is one or more of (rwx). R tells the server the url can be read. W tells the server the url can be overwritten. X tells the server that the url can and should be executed. Access is in addition to normal unix security considerations. For instance a file that can be written to that does not have the W access will have an error returned. The passwdfile is the name of the passwdfile to validate users against. If the passwdfile is given as '.' then the system password file will be used which is /etc/passwd. If no passwdfile is given then no authorization is allowed for anyone. If no users are given then any validated users is authorized, otherwise only the given users are allowed. proxyauth authname authdescription access [passwdfile [users]] proxyauth defines any access authorization to be used for Proxy access authname = Same as auth above authdescription = Same as auth above access = Must be R to allow proxy passwdfile = Same as auth above users = Same as auth above vpath from to [auth [access]] The vpath directive sets up url path translations and authorizations. A requested url that matches from will be translated to to with the given auth and access. If auth does not exist then the url will have no access. If access is not given then the access is taken from the auth record (see above). A '.' in place of the to means that the server should use a translation from another vpath record, but associate the given auth and access with the requested url. A '*' maybe at the end only of the from which is a wildcard match. For example if the from has /AB* then any of /ABCDEF or /AB or /ABmichael will match, but /AD or /a will not. The requested url is first checked against each vpath record until an exact match (meaning url match from and from had no '*') is found or the end of the list. Therefore a wildcard match will match the last from is the list in which it matched. NOTE: if at the beginning of the to field /~user will get translated to the home directory of the given user // wile get translated to the serverroot directory include filename The include directive tells the server to read configuration information from the given filename. NOTE: normally mtype directives are included from another file mtype mimetype extensions The mtype directive tells the server what mimetype to associate with files which have any of the given extensions. If no match is found then the file will be treated as application/octet-stream. NOTE: normally you get mtype directives in included file USAGE: httpd [-v|-t] [configuration-file] The -t tells the server to just parse the configuration file so that you can test it to see if it is the way you want it. You may also pass the name of your configuration file if it is not the default /etc/httpd.conf. The -v option prints the server version and then exits. STARTING: First of all httpd is a server and therefore you will need to start it with tcpd. Tcpd is a program which listens for incoming TCP connections on the passed port and when a connection comes in it forks and starts the given daemon program. Therefore to start httpd you use: tcpd http /usr/local/bin/in.httpd & You will more than likely have this line in your /etc/rc or /etc/rc.net file so that whenever your system is restarted the web server will also be started. The first parameter http is the port that tcpd is going to be listening for connections on. Here http (which should be defined in /etc/services as 80) is the standard port for a web server. The second parameter is the program that tcpd will fork and exec when a connection comes in. The program will then have its stdin and stderr connected to the client Then the web server program will start running with the tcpd program waiting for the next connection. Currently there is no ability to limit the number of simultaneous web servers running. NOTE: At some point I will be adding the ability for httpd to start itself without the need of tcpd. That way httpd will already be in memory and have parsed its configuration file. In Minix 2.0.3 and later versions you may use: daemonize tcpd http /usr/local/bin/in.httpd (daemonize is a shell function defined in /usr/etc/rc which starts programs as daemons). FINAL WORDS I wanted to get the server out as soon as possible so I hurried up and created this document to help out. Hopefully it will HELP more than it HURTS. If anyone is interested in writing man pages for httpd or any of the other network programs please let me know. Michael Temari Michael@TemWare.Com Please note also the SECURITY document in this directory. (asw 2003-07-05)