gadicohen:headers

v0.0.31Published 9 years ago

This package has not had recent updates. Please investigate it's current state before committing to using it in your project.

headers

Access HTTP headers on both server and client (+ clientIP funcs).

For a working demo, see headers.meteor.com.

Copyright (c) 2013-2014 Gadi Cohen, released under the LGPL v3.

For Meteor < 0.6.5, you need to use meteor-headers 0.0.7.

Since 0.0.18, headers are sent down with the initial page access and are available immediately, except when appcache is used. So, please still use the headers.ready() callback. Additionally, proxyCount is deprecated, please use set the HTTP_FORWARDED_COUNT environment variable instead.

Two different types of Headers

When working with Meteor, it's important to understand that there are two different kinds of headers we might be interested in.

  • Initial connection headers -- these are the headers sent to the

initial page access. They are contain a lot of things NOT present in XHR/DDP socket calls. See the difference at headers.meteor.com.

  • DDP headers -- these are the headers sent on the DDP request,

available inside publish and method functions, and accessible directly via this.connection.httpHeaders since Meteor 0.7.1. methodHeaders is still available for now, but might be removed in the future.

See more below.

On the Client

Use headers.get() to access HTTP headers sent by the client. headers.get(key) will retrieve a specific header, e.g. headers.get('host'), whereas headers.get() will return an object with all headers.

Note that all header field names are lowercase. e.g. headers.get('accept-language'), headers.get('x-forwarded-for'), etc (like with ExpressJS req.get).

Headers are available as soon as the package is loaded, i.e. before any of your app code is run, provided appcache is not used. To be safe, run any code that relies on headers.get like this:

1headers.ready(function() {
2	console.log(headers.get());
3});

It's confusing, but headers.ready() called without any arguments is a reactive ready() function which could be used in all the usual places (e.g. with iron-router's waitOn). headers.get() is reactive too. You can use this for example with iron-router to ensure that the headers are available before the first route is executed (no longer required since 0.0.13):

1  Router.map(function() {
2    this.route('/', {
3      template: 'hello',
4      loadingTemplate: 'loading',
5      waitOn: headers,
6      action: function() {
7        // Will only be happen once.  Without the waitOn, you'll see it happens
8        // twice, once before headers are ready, and once after they've arrived.
9        console.log(headers.get());
10      }
11    })
12  });

For more info see the relevant sections in the iron-router docs.

See the notes below on how we retrieve the headers, why it's necessary to do it this way, and also the note about getClientIP().

The following headers are NOT sent back to the client:

  • cookie - dangerous, can be used to bypass httpOnly. access via JS.
  • authorization - uneeded, potentially dangerous
  • user-agent - uneeded, access via navigator.userAgent

On the Server

As above, but you need to pass this as the first argument. e.g. headers.get(this) or headers.get(this, 'host'), and likewise, headers.ready(this). For callbacks / anonymous functions, use var self=this; and call headers.get(self) from within the func.

Available in Meteor methods and publish functions. There is no way to do this for Collection allow/denies unfortunately (but since that relies on a userId, you could use your own logic to set the latest headers per userId)`

Note, headers.get() will always provide the headers for the INITIAL CONNECTION and not the headers for the current connection of the open socket, which is used for publish/methods. Use headers.methodGet(this) to get those headers.

getClientIP() -- freebie

Since a common use for getting the headers to is to check the client's IP address (especially when using a proxy / loadbalancer), I threw in the most common pattern for this.

Usage: headers.getClientIP()

Note, this requires you to set the HTTP_FORWARDED_COUNT environment variable (same as Meteor) with the number of known proxies (that you trust) behind the connection.

Explanation: each proxy in the chain appends to the X-Forwarded-For header, such that if you know the number of proxies, you can work out the initial IP address that the first proxy in the chain specified as the end user's IP address. Thus, even if the end-user sends a request with his own X-Forwarded-For header, you can ignore these fake IPs. In the case where the user's ISP has a transparent proxy, you'll get that proxy's IP... but that remains the only IP that you know for sure is real. See issue #2

methodClientIP() -- server only

Usage: var ip = headers.methodClientIP(this); (inside a Meteor method)

IP address of client connected to the socket over which the Method is running. The same as this.connection.clientAddress in Meteor 0.7.1+.

How and why this all works

Intro: headers have to come from the server. Headers are NOT available from the browser via Javascript. There's a trick to make an XHR request in the browser to get the headers, but note, this does NOT INCLUDE ALL HEADERS (see which headers at headers.meteor.com). Unfortunately the ONLY reliable method, to get all headers with their correct values, that the client sends to the server, is via the server.

  1. On the initial request, all headers are returned on the initial page

load and loaded on the client (injected into the HEAD, before even loading meteor-headers client side code). The headers are also returned with a unique token generated for each request.

  1. When client code is loaded, we use a meteor method to send the above

token back to the server, where we can reassociate the headers from the original request to the current socket. This allows for headers.get() (as opposed to headers.methodGet) to return the correct data, which is kept in the connection session (maintained by livedata).

When appcache is used, we don't get the original HTML page with the headers on it, so we just make a script call to an additional internal script to get them.

Headers not sent to the client

The following headers are filtered out and never sent to the client, either because they represent a security risk or are available on the client anyways (and it would waste bandwidth to send them unnecessarily):

  • cookie - use document.cookie for anything not marked as httpOnly
  • user-agent - use navigator.userAgent
  • authorization

References