https://github.com/frioux/cgid

UCSPI compatible CGI server
https://github.com/frioux/cgid

Last synced: 2 months ago
JSON representation

UCSPI compatible CGI server

• Host: GitHub
• URL: https://github.com/frioux/cgid
• Owner: frioux
• License: apache-2.0
• Created: 2016-02-04T10:34:17.000Z (almost 9 years ago)
• Default Branch: main
• Last Pushed: 2020-06-13T18:02:55.000Z (over 4 years ago)
• Last Synced: 2024-09-18T08:32:57.673Z (3 months ago)
• Language: Rust
• Homepage:
• Size: 17.6 KB
• Stars: 3
• Watchers: 2
• Forks: 1
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# cgid

`cgid` is a [UCSPI](http://cr.yp.to/proto/ucspi.txt) compatible

[CGI](https://www.ietf.org/rfc/rfc3875) server.  It currently supports exactly

one script, though there are plans to support a directory.

## Usage

Here are a couple examples of using `cgid` with `UCSPI`

### [`nosh`](http://homepage.ntlworld.com/jonathan.deboynepollard/Softwares/nosh.html)

```

#!/bin/nosh

tcp-socket-listen 127.0.0.1 6000

tcp-socket-accept --no-delay

cgid

www/cgi-bin/my-cgi-script

```

### [`s6`](http://skarnet.org/software/s6/index.html)

```

#!/bin/execlineb

s6-tcpserver 127.0.0.1 6000

cgid

www/cgi-bin/my-cgi-script

```

(other examples are very welcome)

## Description

`cgid` implements a (hopefully reasonable) subset of the `CGI` protocol.  It

almost surely has glaring missing features, as I have only run a single program

underneath it.  Bug reports and patches are warmly welcome.

## Known Bugs and Limitations

### First response header **MUST** be `Status`

As far as I can tell the CGI specification allows the response headers to be in

any order.  Because all of my applications set the Status first I have not

written the code to buffer the other headers before the Status is set.  As it

stands any headers set before status will be discarded.  I expect to resolve

this when I get a chance.

### `SCRIPT_NAME` is hardcoded to an empty string

If my reading of [`RFC3875`](https://www.ietf.org/rfc/rfc3875) is correctly,

`SCRIPT_NAME` should be which script is being run.  Given that this server runs

exactly one script, I've decided that setting this environment variable is

unimportant.  If I ever support a directory instead of a single script I'll

resolve this.

### Only `HTTP/1.*` is supported

Technically `HTTP/1.0` supports `HTTP/0.9`, which supports responses with no

response code or headers.  This is not supported for simplicity.

### Response `Content-Length` is unsupported

For both performance and simplicity, the response body is streamed from the

application to the client, so the `Content-Length` is unknown and cannot be sent

to the client unless the application set the response header itself.

### Esoteric Error

There is one possible error that is impossible to report to the client

correctly.  Specifically, if there is an error while streaming the response body

to the client from the application, returning a `500 Internal Server Error` is

likely impossible as the applications status has already been set.