Hubbard

Hubbard is a local service designed to seamlessly interact with private, authenticated instances of GitHub: Enterprise.

Supported Protocols

• Git over HTTP
• Raw git assets from a given tree
• Direct download of release assets using the GitHub API

Configuration

Set GITHUB_URL and GITHUB_ACCESS_TOKEN before starting Hubbard.

The Installation section will provide guidelines for configuring a personal access token in GitHub.

The GitHub URL is the URL of the GitHub instance you're working with. If you're working with a private repo on GitHub, the value here would be https://github.com.

You can also control the URL and access token by writing the values to /etc/hubbard/.hubbard.yml.

Start Hubbard with go run hubbard.go.

Installation

First, you will need an access key. Go to the Settings page of the GitHub instance you plan to work with. From there, you should be able to view your Personal Access Tokens:

You can then choose to Generate a New Token:

Select the permissions you believe are applicable. Once you've created the access token, you should record it in a text file or note so that you can use it for configuring Hubbard in the steps below.

OS X

• Install the binary in pkg/darwin-amd64/hubbard into your $PATH
• Configure your instance using hubbard configure

Linux

Hubbard is also designed to work as a service in a Linux environment.

As a superuser, with GITHUB_URL and GITHUB_ACCESS_TOKEN passed into the environment, you should be able to configure hubbard as a service using:

./register-hubbard-service

Windows

We have a binary for Windows, but we are actively seeking collaborators for validating that the binary works, and for correctly registering with Windows.

Usage

Hubbard serves at localhost:41968. For dependency management over the supported protocols, you can now replace the GitHub Enterprise URL you reference with this address. Hubbard will do the work of proxying over your authentication credentials, and, in the case of interacting with release assets, retrieve the appropriate asset via the GitHub API and provide the necessary download context.

This means that installing assets from npm, pip, bundler or any other package management solution that supports compressed flat-file artifacts operates seamlessly.

Development

Make sure you have Glide:

brew install glide

Install your dependencies:

glide install

Run the server:

go run hubbard.go

You should also commit your built binaries:

# Running make will create binaries in ./pkg for each supported platform
make

Why are we committing build artifacts? Many contexts where we want to install Hubbard may not necessarily have access to release assets.The whole idea behind Hubbard is to avoid bad behavior like this in the future. Here, we assume bit of complexity in one place so that this undesirable behavior doesn't propagate to all other projects.

Future work

• Single-line installation of Hubbard on CI
• Add support for YAML-based Viper configurations to manage multiple GH:E instances and authentication credentials by org / repo