url.parse problem with underscore in hostname

[galonsky]

var url = require('url');
var obj = url.parse('http://bucket_name.s3.amazonaws.com/image.jpg');
console.log(obj);

on 0.4.6 (expected output):

{ protocol: 'http:',
  slashes: true,
  host: 'bucket_name.s3.amazonaws.com',
  hostname: '',
  pathname: '/image.jpg',
  href: 'http://bucket_name.s3.amazonaws.com/image.jpg' }

on 0.4.9, 0.5.0, 0.5.1:

{ protocol: 'http:',
  slashes: true,
  host: 'bucket',
  hostname: 'bucket',
  href: 'http://bucket/_name.s3.amazonaws.com/image.jpg',
  pathname: '/_name.s3.amazonaws.com/image.jpg' }

[bnoordhuis]

Won't fix, sorry. Underscores are not allowed in hostnames.

[isaacs]

Hm, it seems like browsers actually allow this, and DNS lookups and everything work as expected. Also, it's a really tiny fix. Reopening.

[isaacs]

@bnoordhuis Care to review? isaacs/node@dcecfc5

[bnoordhuis]

@isaacs Good patch, wrong problem.

[isaacs]

@bnoordhuis Sure, but it's the problem we can actually solve. I don't know how to get the internet to stop supporting URL conventions that aren't allowed by the spec ;)

[isaacs]

For example, this site loads up fine with curl, Chrome, Safari, and node's HTTP client: http://foo_bar_blatz.izs.me/