CAPYBARAS ARE LOVE

Simple Reverse-Proxy written in go.

Listen to a port and redirect any url matching a pattern to another port.

I made this as a solution for hosting multiple services on the same "url" without relying on having a port straight in the URL. Some companies/organizations block any request to a url using a port as one of their security measures.

First parameter must be the path to your json/yaml config file

Example of config.json

{
    "proxy": {
        "port": 80
    },
    "services": [
        {
            "id": "duck",
            "pattern": "^/kwak/",
            "port": 9090
        },
        {
            "id": "stoned",
            "method": "string",
            "pattern": "/jesus",
            "port": 9091
        }
    ]    
}

Example of config.yaml

{
    proxy:
        port: 80
        tls_hosts:
            - localhost

    services:
        -   id: bypasscors
            method: string
            pattern: /bypasscors
            port: 8080

}

Entries in "services" by id:

• "duck" will redirect any request caught on port 80, starting with "/kwak" to the port 9090 using a regex as matching (default) method.
• "stoned" will redirect any request caught on port 80, starting with "/jesus" to the port 9091 using a string as matching method. This method compare the string with the beginning of the URI. It does not try to find the string inside the URI.

Public routes

A service can restrict which routes are publicly reachable with the "routes" key. Without it, every route matching the pattern is served.

services:
    -   id: bypasscors
        method: string
        pattern: /bypasscors
        port: 8080
        routes:
            users: [GET, POST, PATCH]
            books: ["*"]
            books*: [GET]

• Routes are relative to the pattern: users matches /bypasscors/users. The query string is ignored.
• users only matches users, not users/42.
• A trailing * matches the route and everything under it: books* matches books, books/42 and books/42/pages, but not booksellers. books/* only matches what's under books/, and * alone matches every route.
• "*" as a method allows any method. It must be quoted in YAML: an unquoted * is a YAML alias and the config won't load.
• When several routes match, the most specific one decides: an exact route beats a wildcard, and a longer wildcard beats a shorter one. Above, POST /bypasscors/books is allowed by books, while POST /bypasscors/books/42 is refused by books*.
• Paths containing . or .. segments are always refused.
• Anything else gets a 404. Methods aren't implied: list OPTIONS if browsers need CORS preflight on a route.

/ ! \ On Linux (did not try on other system), capybara must be run with sudo if chosen proxy port is under 1024. It will fail otherwise.

This project still needs:

• Tests
• Refacto (so it can be more easily tested)
• Allow config refresh through config file
• More methods of matching besides "string" and "regex"
• Refacto "RemovePattern" behavior to a "RedirectPath" behavior

This project might need:

• Better logging ?