http serving with linux

httpd, cgi-bin and w95

Contents

1 introduction

introduction

after having installed most of the internet services on my p100, i logically wanted to have a full-featured http server, too. not only to serve static http pages to any other computer, but also to engage and learn cgi scripting.

this is mainly to allow database searches via a http interface, os-independant. this is going to be a add-on which many companies will give their clients in the future: simply to know how far the project is, no matter if its a big machine in production or just a parcel sent from A to B.

beware! i only put down the bare bones of scripting here. i even did not install perl.

must be i like bare bones.

installing and configuring httpd

fine. i grabbed httpd.tgz from my suse distribution, added some lines to /etc/inetd.conf, and there i was.

nothing worked.

bullshit!

i tried to read the documentation and found it was in postscript. so why on earth can't they go with some nice ascii files???

as i have not installed gs nor ghostview (nor xwin, this will be after i got through its dedicated client/server concept (making it able to run on any terminal anywhere in the world)), i helpe myself with a line in ~/.mc.ext:

ps grep "(" %f | less

because the lines starting with "(" contain readable output.

soon i figured out that there had to be 4 conf files in /var/httpd/conf: httpd.conf, access.conf, srm.conf, and mime.types.

with the comments provided, httpd.conf was done in 45 secs or so.

srm.conf was faster, i changed nothing. (it was hard for me to understand that the root of my http server was in httpdocs, but the cgi-bin root in cgi-bin, but that's aliases that are kept in srm.conf. perhaps i should have read it before just copying the template.)

access.conf i didn't understand. httpd definitely disliked the syntax. there are some html tags in this file - maybe that's why.

my new access.conf has a length of 0 bytes. THAT WORKS - though i guess i have no directories limited to someone. but security concerns me AFTER anything else is working - so what.

start up!

i put an ugly index.htm (saying "hello") in htdocs, restarted inetd, and lynxed to http://localhost.

it worked, and displayed index.htm.

i lynxed to http://localhost/cgi-bin/date, and it displayed the date.

without the http:// or the localhost/, it wouldn't. bang. i had to set up index.htm with full links.

so good, so far.

cgi-scripting - the bare basics

all right. to get something useful from a cgi script, you must usually pass it some parameters.

it may be simplest to pass them by a user-friendly form. here an example:

oh shit grap the source of this page and carry on with THAT

you should set the action (in line 7) to be sent to metacrawler.com, then this should work.

submitting this form with method=get will pass the parameters (such as general, method,...) as environment variable to the cgi-script named crawler.

i had to set a line "export" in crawler to get all environmets variables displayed to find out that the environment variable with all parameters glued one after another was called "QUERY_STRING". They are separated by & signs, and space are transmitted as &20 (read: hex20, that is dec32).

fumbling around

the advantage is that the variables will always come in the same order. the rest is wild bash-scripting to separate them one from another. the program "tr", which can change &s to, say, LFs, will do great things - if you get it running.

have fun!

17.06.2001

well somewhere on the web i found a functioning cgi script that did NOT use perl to fumble apart the 7 params that are delivered by the script given above. it just used common unix progams as cut, cat, and the like.

i found it to have a response time of 5 secs. so i decided to let it be. there's nothing i must serve to the world, though, that doesn't fit into my table of contents.