GOPHERD/GOPHERLS/GINDEXD(8) System Manager's ManualGOPHERD/GOPHERLS/GINDEXD(8) NAME gopherd, gopherls, gindexd - serve data using the gopher protocol. SYNOPSIS gopherd [-DCIc][-o optionfile] [-L loadavg] [-l logfile] [-u userid] [-U uid] Datadir Port OPTIONS Datadir is a directory where the information for the gopher server resides Port is the socket port to use (or if running under inetd, the port that the server is running under.) Changing the name of the gopherd executable will change how gopherd will behave. If invoked as gopherls , the program will list out the Datadir to the screen, processing all the links and .cap things. If invoked as gindexd, the program can serve up the single index Datadir. We discourage the use of gindexd, since it adds an extra process that must be run. It is included for backwards compatibility. -D Enable copious debugging output. -L If the server is compiled with the LOADRESTRICT option then this option sets the load average at which connections are turned away. -C Disables caching of directory requests. -l Log connections to logfile If a logfile is specified, the server will log the time, host and transaction of each connection. If there isn't a -l option passed to gopherd then no logging is done. -I Specify when running under inetd This is also useful for testing server code changes. -o Specifies an alternative/optional "gopherd.conf" file. See the gopherd.conf(5) man page for more information about the format of this file. -C Turns server side caching off The caching mechanism works as follows: The server checks for a file called .cache when doing a directory lookup. If it finds a relatively new one, it blasts it across the connection, no sorting required. If it doesn't find one, or finds an old one it re-sorts through the directory and makes a new .cache file. -c Do not chroot(2) before processing connections. (Waterlooism) This relies on code that (tries to) ensure(s) that files outside the data directory cannot be retrieved. This has the potential to be a security hole. Use with care, and with the -u or -U options. Also, if you are using it with inetd, be sure to specify the -I flag *first*. This is very useful if you wish to have symbolic links to files outside the data directory available through the gopher server. For example, you might make a symlink to some local man pages, to avoid duplicating the information. If you are running without the -c flag you'll find that certain things might not work right as you expect. These are situations where you need to have some binary executed under the gopherd, or need to have links to files which in reality are outside the gopher data root. Such cases include for example local man-pages etc. Symlinking them is no answer, either a physical link into another directory on same disk, or using a loop-back filesystem (where available) are the ways to do. -u Run as user userid. (Waterlooism) The server is run with reduced permissions (i.e. other than root) which can ensure that only publically readable files are available from the server. This option can be used with or without the -c flag. If you're running a gopher server on top of an anonymous ftp directory you'll want to run with the option "-u ftp" to preserve the same level of security as anonymous ftp. -U Like -u, but takes a numeric uid instead. (Waterlooism) DESCRIPTION Gopherd serves data using the Gopher protocol. The gopherd server understands how to serve text and directory objects. (For indexes see gindexd(8)) The server also supports the creation of "Links". Thus it is possible to represent Telnet sessions, CSO servers, etc. The gopher server also contains enough intelligence to represent mail spool files as directories. ADDING INFORMATION TO THE SERVER The gopher server gets its information from the files and directories in Datadir. By making changes in this directory tree, you will change what the gopher clients see. DEFAULT TRANSALTIONS NOTE: The defaults mentioned in this section apply to the gopherd.conf file that is distributed. If you change this file, then this default behavior will be different. All files/directories that start with a dot, or are named etc, usr, bin, dev, core are ignored or processed specially. The title of each object presented to the client is the filename of the file/directory. Directories in the tree are served as Gopher directory objects. Text files are represented as Gopher text objects. Ulaw sound files are are represented as sound objects. Compressed files are assumed to be text, and have their titles changed to remove the".Z" extension. Mail spool files are represented as directories. When the client attempts to view the mail-spool file, the Subject lines are presented as the title of each mail message within the file. Executable shell scripts are assumed to be of text object type. When the client requests to view the shell-script, the script is run, and the results sent to the client. Files that end with .html are assumed to be WWW hypertext files. They are served up with type 'h'. Files that end with .src are assumed to be WAIS searches. If the server has WAIS indexing compiled in, then these will show up as a search engine/ The server sorts directories it sends out alphabetically. The server does distinguish between upper and lower case when sorting. Consult the gopherd.conf file and the gophed.conf(5) manual page for more information about translations of filename extensions. LINKS You can override the defaults by creating what are known as Links in the Gopher data directory tree. The ability to make links to other hosts is how gopher distributes itself among multiple hosts. There are two different ways to make a link. The first and simplest is to create a link file that contains the data needed by the server. By default all files in the gopher data directory starting with a period are taken to be link files. A link file can contain multiple links. To define a link you need to put five lines in a link file that define the needed characteristics for the document. Here is an example of a link. Name=Cheese Ball Recipes Numb=1 Type=1 Port=150 Path=1/Moo/Cheesy Host=zippy.micro.umn.edu The Name= line is what the user will see when cruising through the database. In this case the name is "Cheese Ball Recipes". The "Type=" defines what kind of document this object is. The following is a list of all the defined types: - -- Ignore this entry (for .cap/XX -files) 0 -- Text File 1 -- Directory 2 -- CSO name server 4 -- Mac HQX file. 5 -- PC binary 7 -- Full Text Index (Gopher menu) 8 -- Telnet Session 9 -- Binary File s -- Sound e -- Event (not in 2.06) I -- Image (other than GIF) M -- MIME multipart/mixed message T -- TN3270 Session c -- Calendar (not in 2.06) g -- GIF image h -- HTML, HyperText Markup Language The "Path=" line contains the selector string that the client will use to retrieve the actual document. The Numb= specifies that this entry should be presented first in the directory list (instead of being alphabetized). The "Numb=" line is optional. If it is present it cannot be the last line of the link. The "Port=" and "Host=" lines specify a fully qualified domain name (FQDN) and a port respectively. You may substitute a plus '+' for these two parameters if you wish. The server will insert the current hostname and the current port when it sees a plus in either of these two fields. An easy way to retrieve links is to use the Curses Gopher Client. By pressing '=' You can get information suitable for inclusion in a link file. OVERRIDING DEFAULTS If you define CAPFILES for SERVEROPTS in Makefile.config then the server looks for a directory called .cap when parsing a directory. The server then checks to see if the .cap directory contains a file with the same name as the file it's parsing. If this file exists then the server will open it for reading. The server parses this file just like a link file. However instead of making a new object, the parameters inside the .cap/ file are used to override any of the server supplied default values. For instance say you wanted to change the Title of a text file for gopher, but don't want to change the filename. You also don't want it alphabetized, instead you want it second in the directory listing. You could make a set-aside file in the .cap directory with the same filename that contained the following lines: Name=New Long Cool Name Numb=2 The replacement (and default) for .cap files are extended link files. The equivilant is to create a file that begins with a dot (.) in the same directory as the file you wish to override. If the name of the file was file-to-change then you could create a file called .names with the following contents Path=./file-to-change Name=New Long Cool Name Numb=2 ADDING COOL LINKS One cool thing you can do with .Links is to add neato services to your gopher server. Adding a link like this: Name=Cool ftp directory Type=1 Path=ftp:hostname@/path/ Host=+ Port=+ Name=Cool ftp file Type=0 Path=ftp:hostname@/path/file Host=+ Port=+ Will allow you to link in any ftp site into your gopher. Make sure that there is a /tmp directory to store the files for the gateway. Note that if you're running without the -c option, you must create a "tmp" directory at the top level of the gopher-data directory. You can easily add a finger site to your gopher server thusly: Name=Finger information Type=0 Path=lindner Host=mudhoney.micro.umn.edu Port=79 Another neat thing you can do is to execute shell scripts: Name=Execed command name Type={a type} Path=exec:"args":/scriptname Host=+ Port=+ This is usually used by other types of gateway scripts. For instance, The first script might take a search and get a few hits. It could then generate "exec" scripts that would retrieve the actual document the hit referred to. Note that the scriptname *must* begin with the magic character "#!/". It also must be executable. HIDING AN ENTRY This kind of trick may be necessary in some cases, when gopherd.conf's "ignore:" tags don't, or can't cover something specific, and thus for object "fred", the overriding .names file entry would be: Type=X Path=./fred by overriding default type to be "X". This kind of hideouts may be usefull, when for some reason there are symlinks (or whatever) in the directory at which the Gopherd looks at, and those entries are not desired to be shown at all. FULL TEXT INDEXES As of the 0.8 release of gopher, full text indices can be accessed directly using gopherd. Just put the indexes someplace in the gopher- data directory and then create a link. For example, assume you have a directory that contains some indexes called "fred" in the top level of your Gopher data directory. Making a WAIS link To make the "fred" directory into a WAIS full text index searcher, add the following lines to .cap/fred Type=7 Path=7/fred/index This will change the type into a full text index (7). The Path= format for a WAIS index is "7{path}/{dbname}". Making a NeXT Digital Librarian link To make the "fred" directory into a NeXT full text index searcher, build an index in the index directory and add the following lines to Type=7 Path=7/fred This will change the type into a full text index (7). The Path= format for a NeXT index is "7{path}". Merging multiple indexes (formerly mindexd) The gopher server can spread an index query to multiple search items, even items on other hosts. This allows one to split indexes into smaller, more managable pieces, and take advantage of a sort of fine- grain parallelism. A multiple search is specified as follows. A file ending with .mindex is placed on the server, inside this file are lines like the following: <host.domain|localhost> <port> <pathname> Here's an example: # #Recipes # localhost 70 7/indexes/otherrecipes ashpool.micro.umn.edu 70 7/indexes/recipes # # computer info joeboy.micro.umn.edu 70 7/indexes/computer The server displays this to the user as a single searchable item. A search to this item will contain the results of searches on the two databases. If you specify "localhost" as the hostname then the server will do the search directly on the local machine in series. Otherwise a copy of the gopher daemon will be forked for each index. Using localhost is very efficient, and should be used when the indexes to be searched are on the local hard disk. GOPHER+ FORMS (a.k.a. ASK BLOCKS) Gopher+ forms allow you to request a variety of data from a person using a Gopher+ client. When the user completes the form, the responses are fed back to the gopher server for processing. Ask blocks are implemented in two parts. The first is a file that defines the form layout. The second is a command script that acts on the responses returned to the server. If you want an ask block named "framis" framis.ask holds the form definition framis holds the script If either is missing, the item is ignored. FORM ITEMS Ask: prompt [<TAB> defaultvalue] displays all text in the prompt up to about 40 chars. and solicits a one-line of input. AskP: prompt [<TAB> defaultvalue] does same but does not echo what you type. AskL: prompt prompts for a multi-lined response. Note: text displays "text" on the screen. Select: prompt displays prompt followed by a "check box". Choose: prompt <TAB> choice <TAB> choice ... displays prompt and list of choices in radio-button format. The first item is the default item for the list. SERVER SCRIPT. First line of script must contain the launching command (e.g. #!/bin/sh ) Otherwise, the server will not act on the script. The script must also have the execute bit turned on. Each response from the form is fed to the script one line at a time. This allows you to simulate user input by just running your script and typing the following items for each line of the FORM code. Ask and AskP returns the user typed text as one line items are ignored. Select returns 1 if selected, 0 otherwise Choose returns the name of the item chosen from the list. AskL sends the number of lines that were typed first, followed by the actual typed multi-line text. The script should read the number and then read the next n lines to get the whole response. If your script contains any "echo" commands or other screen output, that output is returned to the client as a text document. Use this to return feedback to the user, or as a thank-you note. WWW and HTML support. Gopherd can be queried by WWW clients in the hypertext mode. The server does not currently cache the HTML pages it makes. You can include HTML stuff that will appear above the directory by making a file called “.about.html” The gopher server will send this before the menu. References to Gopherd-WWW pages are as follows: gopher://hostname.domainname:port/hh/gopherpath EXAMPLES /usr/local/etc/gopherd /home/gopher-data 70 This command will serve data from the directory /home/gopher-data on port 70 The gopher daemon forks off and will then start accepting connections. You can easily test the server by using telnet to connect to the port specified on the command line. Once connected, type return. A list of things in the gopher-data directory should be returned. For instance let's say that the server was started on the machine mudhoney. To test it you'd do the following: telnet mudhoney 70 Here is an example of running the gopherd server from inetd. This example assumes that you have the files /etc/inetd.conf and /etc/services. Other machines and architectures may vary. Assume that we have a directory /gopher-data and we want to run at the standard port for gopher (70). In /etc/services add the following line: gopher 70/tcp In /etc/inetd.conf add the following line: gopher stream tcp nowait root /etc/gopherd gopherd -I /datadir 70 AIX 3.2 users take note; when adding gopher to /etc/inetd.conf, the command string must be 50 characters of less. This is because under AIX the inetd information is stored in the ODM object InetServ in /etc/objrepos. If you use more than 50 characters, the command will just be truncated when /etc/services and /etc/inetd.conf get imported into /etc/objrepos/InetServ by the inetimp command. For reference you can get the description of the InetServ object class by going 'cd /etc/objrepos' and 'odmshow InetServ'. To see the contents of the InetServ object itself, go into /etc/objrepos and do 'odmget InetServ'. FULL-TEXT INDEX EXAMPLES An full-text index is used to rapidly find data in a set of files. The first step in making a gopher server with indexes is to build the index on your data files. For some canned indexing scripts look in the misc/indexing directory. They might be more intelligible than these instructions. When using the NeXT digital librarian this command is ixBuild , when using the WAIS software this command is waisindex Okay, how about a real world example? Suppose that you have a directory structure that contains your grandmother's favorite recipes located in /home/mudhoney/recipes/. Also assume that there are subdirectories for cakes, pies, and stews i.e. : % pwd /home/mudhoney/recipes % ls cakes/ pies/ stews/ In this example we will be running a gopher server on the directory /home/mudhoney/recipes. Building the index with the NeXT and ixBuild. Go into the directory you want as the root level of your index. If you want to index all of the recipes you'd type the following: cd /home/mudhoney/recipes mkdir .index ixBuild -Vv This will make an index in the file: /home/mudhoney/recipes/.index/index.ixif If you wanted to just index the pies subdirectory you would do the following: cd /home/mudhoney/recipes/pies mkdir ../.index ixBuild -V -i../.index/index.ixif or... cd /home/mudhoney/recipes/pies mkdir .index ixBuild -V -i.index/index.ixif pies It's important that the filenames that are generated by ixBuild have the same format that is given by the gopher server! You can easily test this by using ixFind path. Also make sure that when you're indexing a subdirectory that the previous directories up to the gopher data directory get added to the path by ixBuild. Building the Index with waisindex With waisindex it isn't necessary to be in the root directory of the gopher server. The WAIS indexer stores filenames with their absolute paths, this causes problems later. If you want to index the whole recipe collection in /home/mudhoney/recipes you'd do the following: mkdir /home/mudhoney/recipes/.index cd /home/mudhoney/recipes/.index waisindex -r /home/mudhoney/recipes The waisindex program will create a bunch of files starting with index. These files together comprise an index to your data. If you wanted to just index the pies subdirectory you would do the following: mkdir /home/mudhoney/recipes/.index cd /home/mudhoney/recipes/.index waisindex -r /home/mudhoney/recipes/pies Linking in the Sample indexes Both of these indexes are in a directory that starts with a period. Thus the server will ignore these directories when sending out information. So now you need to make a link to these indexes so people can start searching them. For the NeXT you would put this in a file /home/mudhoney/recipes/.Link Type=7 Name=Recipe Index Port=+ Host=+ Path=7/ For a WAIS based gopher server you would put the following into the file /home/mudhoney/recipes/.Link Type=7 Name=Recipe Index Host=+ Port=+ Path=7/.index/index Changing the defaults for full text indexes... You may want to change what gets returned for a full text index. Especially if you are storing indexes on a machine that doesn't have the actual data. In this case you can use what is known as a "hostdata" file to change what the server returns given a full-text search. The server will look for a file called "<indexname>.hostdata" to get information from. If it doesn't find this file it will try just "hostdata". A hostdata file consists of a text file with three lines in the following order: <hostname> <portnum> <path to strip> These fields are used to override the defaults. For instance, the following hostdata file would substitute the hostname "goober.micro.umn.edu" for the current server host, the port 8000 for the current server port, and strip off the path /home/goober from the pathname entries in the index: goober.micro.umn.edu 8000 /home/goober SEE ALSO gopher(1); gopherd.conf(5); The Internet Gopher Protocol, March 1993, RFC 1436 Local GOPHERD/GOPHERLS/GINDEXD(8)