~ushin/hyperdrive.el

P2P filesystem in Emacs
Tidy: (hyperdrive--gateway-ready-p) Remove unused function
Docs (hyperdrive.el): Installation instructions in Commentary
Docs (README.md): Clarify start/stop instructions

refs

master
browse  log 

clone

read-only
https://git.sr.ht/~ushin/hyperdrive.el
read/write
git@git.sr.ht:~ushin/hyperdrive.el

You can also use your local clone with git send-email.

#hyperdrive.el - P2P filesystem in Emacs

Hyperdrive is a secure, P2P, real-time, local-first, versioned filesystem designed for easy peer-to-peer file sharing. hyperdrive.el is an independent project built by USHIN which provides an Emacs interface for managing hyperdrives.

#Installation

hyperdrive.el is available on MELPA. Once you've set up MELPA, you can run M-x package-install then enter hyperdrive.

#hyper-gateway

hyperdrive.el relies on hyper-gateway for talking to the hypercore network. Download or compile the binary then either run it manually or install it as a systemd service.

#Usage

Be careful what you publish! Anyone with your public key can download your published files from you, your friend, or anyone else who has them.

#Start/stop the gateway

To connect with peers, you'll need to start hyper-gateway. If you install hyper-gateway as a SystemD service, you can connect and disconnect from the network with M-x hyperdrive-start and M-x hyperdrive-stop. Otherwise, follow these instruction to run hyper-gateway manually.

#Create a hyperdrive

You can have multiple hyperdrives, each one containing its own set of files. Run M-x hyperdrive-new then type in a seed to create a new hyperdrive. That seed will be combined with your secret master key, which is generated for you by hyper-gateway, to produce a public key that uniquely identifies that hyperdrive. hyperdrive-new is idempotent since the same seed will always produce the same public key. For this reason, a hyperdrive's seed cannot be changed.

#Open a hyperdrive

You can open a hyperdrive folder or file by pasting in a hyper:// URL after M-x hyperdrive-open-url. Alternatively, M-x hyperdrive-find-file remembers hyperdrives you have already created or visited. It will prompt you to select a known hyperdrive by its public key or seed and then enter a path inside that hyperdrive.

The following keybindings are available inside the directory view by default:

  • n and p move between entries
  • RET opens file or directory at point
  • ^ goes up to the parent directory
  • D deletes the file or directory (recursively) at point
  • w copies the URL of the file or directory at point

#Write to a hyperdrive

You can write a buffer to a hyperdrive with hyperdrive-write-buffer, which will prompt you for one of hyperdrives you have created as well as the path in that hyperdrive where you want to store the file. If you are editing an existing hyperdrive file, hyperdrive-save-buffer will silently update the current hyperdrive entry with the new content.

In addition to copying the URL at point in the directory view, you can run hyperdrive-copy-url to copy the URL of the current hyperdrive file or directory. If the current file is an org-mode file, org-store-link will store a link to the hyperdrive file, and if point is inside a heading, its CUSTOM_ID, ID, or heading text will be appended to the stored URL.

#Share a hyperdrive

Only you can load one of your created hyperdrives by its seed. When sharing a hyperdrive with someone else, you will need to copy its full URL. Peers can load your hyperdrive files directly from your computer or from other peers who previously loaded those files.

It is possible to use DNSLink to link to a hyperdrive with a domain name instead of a public key, like hyper://example.org/path/to/file. Create a TXT record at _dnslink.example.org with the contents /hyper/<public-key> (no trailing slash). Note: relying on DNS adds another point of centralization, reducing the durability of your link. hyperdrive.el somewhat mitigates this issue by remembering which public key the DNS record resolved to, so that peers can use the stored public key itself for subsequent connections.

#Upload files from your filesystem

TODO: Uploading files from the filesystem has not been reimplemented since the upgrade to the latest version of hyper-gateway.

hyperdrive-upload-files lets you upload files from your filesystem to a hyperdrive. It accepts a seed, a relative-dir, the directory on your filesystem which will be the top-level or root directory of the hyperdrive, and the list of files to be shared in the hyperdrive. files can either be a list of filepaths or a function which returns a list of filepaths.

In order to upload files to a hyperdrive, you must first create it with hyperdrive-create-namespace.

#Upload a whole directory

my/hyperdrive-upload-files-foo uploads everything inside of ~/public/:

(defun my/hyperdrive-upload-files-foo ()
  "Upload all files inside of \"~/public/\" to hyperdrive with seed \"foo\"."
  (interactive)
  (hyperdrive-upload-files "foo" "~/public/"
                           (lambda () (directory-files-recursively "~/public/" ""))))
#Upload files by tag

my/hyperdrive-upload-files-bar uploads all files inside ~/org/ which have been tagged as "public" using Protesilaos Stavrou's Denote file-naming scheme.

(defun my/hyperdrive-upload-files-bar ()
  "Upload all files tagged \"public\" inside of \"~/org/\" to hyperdrive \"bar\"."
  (interactive)
  (hyperdrive-upload-files "foo" "~/"
                           (lambda () (directory-files-recursively "~/org/" ".*_public.*"))))

Alternatively, you could select files by tag with Karl Voit's filetags. Either way allows for a "non-splitting" approach where public and private files exist in the same directory. You can write any function you like to determine which files to share!

#Concepts

#Hyper-gateway

Hyper-gateway handles interactions with hyperdrive under the hood, and it runs a local HTTP server which accepts HTTP requests. In hyperdrive.el, P2P interactions consist mostly of, e.g., GET requests to download files and PUT requests to write files to a hyperdrive.

#Sparse replication

Hyperdrive is sparsely replicated, meaning that peers can download particular files from a hyperdrive without having to get the whole drive. This reduces both load times and disk usage.

#Hyperdrive entries

Instead of files and folders, Hyperdrive has entries and entry prefixes. It's possible for a "file" entry and a "folder" entry prefix to have the same name, like hyper://<public-key>/path/to/ (a "directory") and hyper://<public-key>/path/to (a "file"). In this case, the directory listing for hyper://<public-key>/path/ would display the to entry but not the to/ entry prefix. Because entry prefixes only exist when they prefix an entry, deleting the last "file" in a "directory" causes the "directory" to disappear as well. When a hyperdrive "file" or "directory" is not found, hyperdrive.el prompts to go to the parent "directory."

#Bugs and Patches

Bugs can be submitted to the ushin issue tracker. Patches, comments or questions can be submitted to the ushin public inbox.

#Acknowledgments

Adam Porter for rewriting hyperdrive.el and for his work on plz.el.

Mauve Signweaver for their guidance into the world of p2p as well as the development of hyper-gateway.

Karl Voit for his feedback, especially the suggestion that we allow for a non-splitting approach for uploading files from the filesystem.