bhpack

Pack binary header fields in HPACK more efficiently

Usage

Encode binary data as strings that are more efficiently encoded in HTTP/2.

var bhpack = require('bhpack');
// For cookies
var cookies = {
  a: bhpack.cookie.encode(valueOfA)
};
// For URLs
var url = 'https://example.com/' +
          bhpack.uri.encode(pathComponent) +
          '?q=' + bhpack.query.encode(qParam);

Decode the resulting values again.

var cookieA = bhpack.cookie.decode(encodedString);
var url = require('url').parse(url, true);
var pathComponent = bhpack.uri.decode(url.pathname.substring(1));
var qParam = bhpack.query.decode(url.query.q);

Performance

It's a few percentage points. That's not a big saving unless your bandwidth costs are dominant.

I could implement this with arithmetic or range coding, which would shave this down slightly, but that hasn't happened yet, and probably won't.

Here's what node compare.js 100000 20 returned one time. That is, what the compression is like for 100,000 random 20 byte sequences for the different modes.

100000 random sequences of lengths:     20
Average sizes: min/ave/max (size compared against Base64+Huffman)
Raw Huffman coding (invalid):           25/48.20803/64 (203.13%)
Base 64 (no Huffman):                   27/27/27 (113.77%)
Base 64 with Huffman:                   20/23.73286/27 (100%)
bhpack cookie with Huffman:             19/23.6052/30 (99.46%)
bhpack URI safe with Huffman:           20/22.98342/28 (96.84%)
bhpack URI query with Huffman:          20/23.01238/28 (96.96%)

FAQ

Why?

Because HPACK is kind to URI-safe base64, but not as kind as it could be.

Why not?

You don't want to do this for security reasons. Huffman encoding reveals a tiny bit about your data based on the length that it encodes to. This can be slightly worse.