aka Roots for SproutCore. Dobby, the sweet little magical server that pushes.
Python
Switch branches/tags
Nothing to show
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
clients
dobby
.gitignore
.gitmodules
Dedication
README.md
dobby.py

README.md

Dobby (codename for Roots) is a lightweight push server. Its core is Thestral, a simple protocol-independent interface for pushing information around.

Quick+Simple Part

At its most simple, Dobby is a lightweight but flexible event-based system which includes a Comet server component.

Warning

First, this walkthrough really isn't thorough enough. Now, the other issues: a) Dobby is under somewhat heavy development. b) You should not use in production without understanding a little of the innards—which is not difficult to do, given that it is extremely simple. There are some security issues, such as ports that would likely need blocking.

Conclusion: Use Dobby if you a) just want something quick and not necessarily permanent, or b) want to actually learn Dobby's code somewhat fully and help out. :)

How you use

Typically, an application using Dobby will have three parts:

  • Dobby
  • Back-end (Django, Rails, etc.)
  • Front-end (SproutCore)

The Workflow

In short:

  • Pomona (SproutCore framework for Dobby) connects to Dobby.
  • Dobby sends Pomona an ID.
  • Pomona calls two URLs given two it to connect or disconnect from paths/channels.
    • URLs are implemented in back-end (Django, etc).
    • The ID is sent with these requests.
  • Back-end tells Dobby to connect the client to the path.
    • Back-end can do any authorization it wants here, however it wants. If it doesn't want to allow the connection, it can just not tell Dobby.
  • Back-end sends updates to Dobby, whenever it pleases.

Pomona is quite nice in that it keeps track of connections for you, so will automatically reconnect to anything should the connection be dropped.

How to Implement-Client Side

The client side should be implemented using Pomona, the SproutCore framework for Dobby.

TODO: Write here how to clone Pomona and set it up as a framework. You may very well already know how to do that anyway...

Typically, you set up the Dobby connection in whatever Data Source you use. It is really very simple. For example, you may have the following init function:

	init: function(){
		// by default, tries to go through "/comet/", so you need
		// to proxy using sc-server during development.
		this.firenze = Pomona.Firenze.create({
			connectUrl: "/server/connect/%@",  // the paths to use to connect+disconnect
			disconnectUrl: "/server/disconnect/%@"
		});
		
		// connect paths to functions
		this.firenze.connect("contacts", this, "contactReceived");
	},
	
	contactReceived: function(path, message) {
		if (message.trim() === "") return; // first message sent—to confirm connection—is blank.
		
		// read the data
		var data = JSON.parse(message);
		
		// loada the record
		Contacts.store.loadRecords(Contacts.Group, [data]);
	}

Can't get much simpler than that!

How to Implement: Back-end with Django

First, you need to add the "cornelius" package to your Django project. git clone git://github.com/ialexi/cornelius.git

Now, you have to enable connection and disconnection. Some lines like this in urls.py should do it: (r'^contacts/connect/(?P<uid>[^\s]+)$', "contacts.views.connect"), (r'^contacts/disconnect/(?P<uid>[^\s]+)$', "contacts.views.connect")

In your views (contacts/views.py in the above example), you would add some views. Note that, while the following works, I need to update it to use cornelius.dudley, which is vastly superior, oddly enough, to "imperio." # These don't do ANY authentication. You can just check the paths received. # The paths are an array of strings, as you can see. def connect(request, uid): paths = json.loads(request.raw_post_data) # Unfortunately, this will send out a request for each. Optimizations would be nice... for p in paths: cornelius.imperio.connect(uid, p) return HttpResponse("{sent:true}", mimetype="application/json")

	def disconnect(request, uid):
		paths = json.loads(request.raw_post_data)
		for p in paths:
			cornelius.imperio.disconnect(uid, p)
		return HttpResponse("{sent:true}", mimetype="application/json")

You will, of course, also need to include json and cornelius: import cornelius.imperio try: import simplejson as json # For <= 2.5 except ImportError: # for 2.6+ import json

And now, you need to actually send messages. The easiest way is to add some post-save hooks. I add mine in models.py: # Comet alerters def contact_saved(sender, **kwargs): try: instance = kwargs["instance"] cornelius.imperio.update("contacts", json.dumps(instance.toRaw())) except: pass

	post_save.connect(contact_saved, sender=Contact)

And that's it.

How Dobby Works

Dobby is centered around a simple interface called "Thestral."

Thestral is so simple that each instance only has two things: an id and an "update" function that takes three arguments: a sender, a path, and a message.

The Thestral interface is the model that all message receivers fill. There are many kinds of message receivers; for instance, dispatchers (such as "Pig" in "Owl.py") relate paths to other thestrals. They do so by id, and ids are managed by Dolores, the root message receiver.

How do connections between paths and other thestrals get made? Well, by specially formed updates: update("::connect", "id->path")

The long-polling server ("Firenze") is a server that creates Firenze instances (which are Thestral implementors) for each connection. It registers them with Dolores (the root) so they get an id, and then they can be connected by anything that can talk to the dispatcher.