TSWS, A Totally Simple Web Server in Bash and Socat or alternatively Netcat.
sudo apt-get install socat
git clone https://github.com/dfletcher/tsws.git
cd tsws
./tsws 127.0.0.1 8080 &
chromium http://127.0.0.1:8080
Socat (socat
) is recommended over Netcat (nc
) and Socat is required for
correct operation on the Cygwin platform. The script first checks for socat
and uses that if available. If not, it checks for nc
and uses that as a
fallback solution.
There are a lot of variations of the nc
program around and it might not work.
If nc
complains about a missing -k option, you could try calling the script in
a loop to emulate the operation:
while : ; do ./tsws localhost 8080; done
This will not work well on Cygwin or other Windows based Bash distributions, the delay of launching an executable will cause too much downtime, the browser will often be trying to connect while this is looping and will get no response because it is not listening for connections for an extended period of time.
TSWS_LIBRARY=1 . tsws
declare www_index_Content_Type="text/html; charset=utf-8"
function www_index {
echo "<html><body>My index is better than the default.</body></html>"
}
tsws 127.0.0.1 8080
TSWS_ROOT="/var/www/html" ./tsws
When serving files this way, the server first looks to see if a file exists $TSWS_ROOT/$URL_PATH. If it does, and the file is actually inside TSWS_ROOT or a subdirectory, it will be served up. If the file exists but is not underneath TSWS_ROOT, a 500 error occurs.
When serving files from disk, file -ib
is used to detect the Content-Type.
TSWS_ROOT is optional, if empty no files will be served from disk and all content is in Bash function form.
If the file from the url path does not exist, or TSWS_ROOT is not set, the
server looks for a function. Any characters other than [a-z|A-Z|0-9] will be
replaced with an underscore. The function name is prefixed with "www". The root
path in this scheme translates to "www_", so this is replaced by "www_index" in
order to have a nicer function name for the site index page. The function
should print text or binary content using cat
, echo
, printf
or similar.
When a function is used to serve content, the Content-Type is declared in a variable with the same name as the function with a suffix "_Content_Type".
See the library example above for an example of a content serving function and related _Content_Type variable.
Have a look at the source code for many more details and discussion, including theory of operation and creating dynamic content.
-
!GAPING SECURITY HOLE WARNING! Do not trust the input. An attempt is made to scrub requests, it may not be 100% complete (this project is brand-spankin' new). The core engine looks for bash functions named similarly to the path. These are prefixed with "www_". Think carefully when writing functions named similarly, ensure that values coming from the network cannot be evaluated.
-
All HTTP caching is ignored keep things simple.
-
Parsing CGI variables off URLs is weak. TODO find a decent way to do it that doesn't baloon the example code.
-
Iterative development is awkward. Altering functions while bash is running confuses it, and the server must be restarted to see changes.
-
(Netcat only) this setup almost certainly has issues with concurrency. It is useful for a single local user or demonstration purposes, but please do not use it with Netcat in any kind of production web server connected to the internet.
-
(Netcat only) Cygwin doesn't work so great. Particularly, some requests just fail inexplicably with nc never supplying a request or supplying a massively delayed one.
-
(Netcat only) It is easily confused, especially if multiple browsers or tabs are pointed to it at the same time. Requests might get dropped and the order can get out of whack, sending the wrong file for a request.
-
POST, file upload.
-
Deprecate
nc
, remove docs. -
Improved URL parsing.
-
Stuff request headers into map available in content functions.
-
Detect envirornment, suggest course of action for missing deps.
-
Improve code flow, _tsws_response is overly long and hard to read.
-
It runs slow, investigate ways to improve performance.
-
Bug fixes reported in github issue queue and via pull requests.