File Name
gopherd server in versions 1.0 and above. You probably won't need
this code.
In this distribution is a simple and relatively (ha!) readable
Perl script called g2ftpd. The script implements an internet
Gopher to FTP gateway. The gopher client connects to the g2ftpd
daemon and tells it what the name of the ftp site is, (and perhaps
what file it wishes to retrieve from there). The daemon connects
to the remote host, retrieves a directory listing (or file) via
anonymous ftp, and returns it to the gopher client. Of course
all this happens without any new knowledge by the gopher client.
Running the script on a UNIX box gives folks running gopher clients
access to all that good stuff available out there for anonymous ftp.
The script assumes a nice healthy unixy ftp server at the ftp end...
that is to say it is probably not going to work with funky IBM or
TOPS-20 ftp implementations. This is left as an exercise for the
hacker... er... reader.
The script is designed to be run by inetd as a TCP daemon.
The easiest way to explain how to set it up and how it works
is to use an example.
Let us assume that we want to run a gopher to FTP gateway on
a machine called hell.micro.umn.edu.
1. You need to have Perl available on the machine. If you don't you
can ftp the sources from a variety of anonymous ftp sites including
uunet.uu.net
jpl-devvax.jpl.nasa.gov
tut.cis.ohio-state.edu
2. Edit the first few, (marked) lines of the g2ftpd file for your
local configuration; in this case change the marked lines as follows:
#----Stuff here must be customized for your machine----
$myName = "hell.micro.umn.edu"; #host full domain name
$myPort = "7997"; #port at which inetd will listen
$ftp = "/usr/ucb/ftp"; #whereever on your box this lives
#----end local customizations-------
3. Become root and place the gtoftpd file someplace nice (like
/usr/local/bin or /usr/etc or wherever you place stuff like this);
assume we put it in /usr/local/bin. Make it owned by root and
executable:
chown root g2ftpd
chmod 755 g2ftpd
4. Update /etc/services by adding the following line to the
/etc/services file (note it's tab between g2ftpd and 7997):
g2ftpd 7997/tcp
Nothing sacred about 7997. Run it at 6666 if you like...
For SUNs running yp, you'll also want to do a make services:
cd /var/yp
make services
5. Now update /etc/inetd.conf (for BSD-ish systems) or
/etc/servers (on Ultrix or A/UX or others) depending on what you
have.
For /etc/inetd.conf, add a line:
g2ftpd stream tcp nowait root /usr/local/bin/g2ftpd g2ftpd
For /etc/servers, add a line:
g2ftpd tcp /usr/local/bin/g2ftpd
Note tabs between fields.
6. Kill and restart the inetd daemon whatever the prescribed way
to do that on your machine.
7. You can confirm that g2ftpd is running now at port 7997 by
telneting to hell.micro.umn.edu at port 7997. Type a hostname at
it (eg: boombox.micro.umn.edu), and it should respond by returning
the anonymous ftp directories available at boombox.micro.umn.edu.
8. Now make some links from a regular gopher server to the g2ftpd
service. On a unix server, the links could look like this (on a
Mac server, do the functional equivalent using Gopher's Helper):
Name=Sumex-aim: Large collection of public domain Mac software.
Type=1
Port=7997
Path=sumex-aim.stanford.edu@info-mac/
Host=hell.micro.umn.edu
Name=boombox: Home of POPmail and Gopher Central.
Type=1
Port=7997
Path=boombox.micro.umn.edu@pub/
Host=hell.micro.umn.edu
And you're all set.
What happens is that the user sees the full Name (as always), say
"boombox: Home of POPmail and Gopher Central". It would appear as
a directory (or folder icon on a Mac). If selected (double-clicked
on a Mac), the client would open a connection to hell.micro.umn.edu
at port 7997. The gtoftpd daemon would accept. The client would
then send it the selector string "boombox.micro.umn.edu@pub/". The
daemon would interpret this as "Go out to boombox.micro.umn.edu
via anonymous ftp and get a listing of the pub directory." The
daemon would return this listing to the client in nice gopher
format. And everything else happens by magic as you would expect.
To the client it looks just like any other gopher-type directory.
The script does assume that files ending in .hqx are BinHex files
and gives them the type 4. It also makes rash assumptions about the
plethora of DOS binary files (.zip et. al.), and maps them all
to a type 5. Type 5 used to be "IBM BinHex" in the protocol notes,
but since nobody really uses that, we've decided to map all these
pesky binaries into type 5. The client can decide what to do with
them based on the .xxx extension that most seem to carry. Finally,
it gives tar and Z files a filetype of 9. And 9 is a new gopher
item type (actually the only one we've allocated since we set up
standards for the protocol. All other files are given normal
type 0: that is text.
If your client is capable of handling binary types then it can
allow type 5 and/or type 9 in directory listings. It should be
prepared to do things based just on the extensions (.xxx) of the
files that the user requests. In particular, it will need to
just read from the connection until the server closes the connection
when sending a binary file. So there will be no concluding period,
and obviously lines are meaningless.
WATCH OUT: The script has no way of knowing for sure that a file at an
ftp site is binary (other than educated guesses based on the suffix of
the filename), so when it presents gopherized listings, it must show
these files too. When a client attempts to retrieve a file, the script
can peek into the file and make a much more educated guess as to
whether it contains text. So it should prevent a client from
transferring totally gonzo binary. I can hear the unix folks, the GIF
folks, the PICT and TIFF folks and all the other special interest
binary folks screaming.... :-) Yes, we need to do add some types... but
we'd like to keep them as few as possible.
It would probably be wise to place caution ReadMe files out there for
the users. Also, if possible good gopher administrators would place
the link all the way inside the "pub" or whatever directory at the
server end. We really don't want users fumbling around inside the bin
or etc directories and attempting to suck down binary files.
Once more: the binary question is still open and we need to resolve
that one yet.
All credit for this great and simple idea should go to Craig Rice
(cdr@stolaf.edu). Thanks Craig.
FXA.