fibridge-proxy-rs

Stream huge files out of your browser without having to upload.
11 Sep 2019491

Live demo instance running at https://fbrg.xyz. Feel free to use it but please try not to overload it (ie lots of video streaming, etc). It doesn't have much memory.

Rationale and introduction available in the blog post.

The point of this is to allow your browser to "host" files which can be streamed over HTTP. This requires a proxy server to handle the HTTP requests and forward them to the browser over websockets.

Why would this be useful? If the user has a very large file (genomic data files can easily be in the 20GB-200GB range), and you want to make ranged requests to that file (ie only download specific chunks) as though it were hosted on a normal server, this will allow that. In iobio we use this to allow our backend system to access a user's local files for processing by backend tools such as samtools.

Example usage

First start up the proxy server. We'll assume it's publicly available at example.com. You can either download it from the release page, or see below for instructions on building.

Then start it up:

./fibridge-proxy-rs --port 9001

A few useful options:

sudo ./fibridge-proxy-rs --host fbrg.xyz --port 80 --ip-address 172.xxx.x.x --key keyfile.pem --cert certfile.pem --secure-port 443

Create a hoster object in the browser and host a file (see this page for information about the fibridge-host library):

const fibridge = require('fibridge-host');
fibridge.createHoster({ proxyAddress: 'example.com', port: 9001, secure: false }).then((hoster) => {

  const file = new File(["Hi there"], "file.txt", {
    type: "text/plain",
  });

  const path = '/' + file.name;

  hoster.hostFile(path, file);

  // This is necessary because fibridge automatically generates a randomized
  // hoster id which is included in the path.
  const fullPath = hoster.getHostedPath(path);
  const url = `http://example.com:9001${fullPath}`;

  console.log(url);
});

That will print out a URL for the hosted file. You can then retrieve the file using any http client:

curl example.com:9001/<hoster-uuid>/file.txt
Hi there

Ranged requests work too:

curl -H "Range: bytes=0-2" example.com:9001/<hoster-uuid>/file.txt
Hi

Building

In order to build from source, you'll first need rust installed. The proxy currently expects the GUI repo to be available in the same directory, like this:

fibridge/
  fibridge-proxy-rs/
  fibridge-gui-js/

See fibridge-gui-js for instructions on building the GUI.

Once the GUI is built, run:

cargo build --release

If all goes well you should end up with a binary in fibridge-proxy-rs/target/release.

Other implementations

There is an API-compatible JavaScript (Node) implementation of the proxy server available here.

API

I still need to document the API, but it's pretty simple. It uses omni-rpc, which is itself built on top of omnistreams.