I think the js.io apis changes from jsio.logging.getLogger to
jsio.logging.get()

More importantly, loggers are in each scope by default, so you can
probably just delete whatever offending getLogger line is there.

Also, you should do any compilations against the 3.2 tag of js.io --
things are changing rapidly in trunk.

On Fri, Aug 27, 2010 at 12:17 PM, Erik Anderson <
Orbited2 supports the same use case that Orbited supports -- a tcp proxy.
Orbited2 is an architectural drop-in replacement for Orbited.
Orbited2 has two basic modes
1) Exposing a WebSocket api to the browser, and sending WebSocket data (with
the WS protocol) to a back-end server.

Trying to wrap my head around this... I think my use case is still
that of using TCPSockets, in which case WebSockets becomes an
infinitely more efficient transport protocol, but the daemon still
maintains its purpose as a transport proxy, at least when you're
dealing with a service that doesn't speak native WebSocket (or

It's getting called the first time by CometTransport's close method,
in base.py. But that gets called after every write. Shouldn't
connectionLost be called only when the client disconnects?

Here's the stack trace from pdb:
(Pdb) w
/usr/local/bin/orbited(8)<modu le>()
-> load_entry_point('orbited==0.7 .10', 'console_scripts', 'orbited')()

I've seen this as well (i.e IRC closing by timeout and Orbited not
closing). I think this is to due with the issues with the CSP protocol
not closing properly (also causing the issues on
[link])

mcarter, I know your busy, but can you check

Not sure if that is quite right... if you look at TCPResource.render (this
is what gets called each time a client attempts to handshake) you see where
an instance of a TCPConnectionResource is created. There should always be
one of these created per connection.
Perhaps the lostTriggered is being set improperly elsewhere. I would think

I'm having trouble getting orbited to kill the connection it's
proxying.

I'm using it to run IRC through a web browser. When the user
disconnects, I would expect the IRC server to show the "remote host
closed the connection" message, but instead I get ping timeout. So
the code in proxy.py (in connectionLost) isn't killing the connection.

Sounds about right. If tornado is a websocket server, then just use
Orbited2.js to put a websocket in IE, etc. and you're good to go.
- Michael Carter

So for instance I could do the following with Orbited2:

Deploy Orbited2, and use the Javascript library Orbited2.js and use
Websockets in any browser (including IE?).
Write a Tornado websocket server, which could deal with authentication
from the incoming request.
Once authenticated my Tornado setup could subscribe to a AMQP message

Hi,
there's a well-known bug with closing browser [link]
What action precedes this exception?

I keep seeing this which stops my server from functioning properly.
What is this error? How do I fix this?

08/23/10 23:18:14:777 DEBUG orbited.transports.base.CometT ransport
calling finish
Traceback (most recent call last):
File "/usr/local/bin/orbited", line 9, in <module>
load_entry_point('orbited==0.7 .10', 'console_scripts', 'orbited')

I would recommend using nginx to proxy connections to
[link] to your actual Orbited port.
Something like:
server {
listen 9000;
server_name localhost;
location /some_path {
# Orbited listens at port 9001
proxy_pass [link]
# Good settings for Orbited

I used to use IM+ on my iphone. Pretty sleek!

maybe you can use URL rewrite
在 2010-8-12 下午7:07，"zagor" <luca.z...@gmail.com>编写：
Hi,
I wonder if it's possible to bind Orbited to respond to connection on
suburl [link] .
In this way the managing by javascript is more comfortable for me.
Have you some ideas?
Luca

Hi Michael,
IM+ uses our proprietary backend service developed for about 7 years
in Java(and some C). It was originally built for our mobile clients.
Some of our competitors use libpurple to support these protocols.
Dmitry.
On Thu, Aug 12, 2010 at 11:59 PM, Michael Carter

Neat; what library are you using to implement all those protocols?

Hi,
just wanted to share my recent work done using Orbited as a comet
backend with TCPSocket on the client side.

So here it is [link]
If there are be any tech questions just ask :)

Dmitry.

Hi,
I wonder if it's possible to bind Orbited to respond to connection on
suburl [link] .
In this way the managing by javascript is more comfortable for me.
Have you some ideas?

Luca

Just a follow up for anyone reading this post later on (Google etc.)

Memory leaks was defined as CSP not closing due to improper timeouts.
This lead to that neither CSP Session or Proxy.Incoming were cleaned.

I have an ugly fix for this in place and mcarter has written a
proposed fix (which is currently waiting on

I agree on the cleaning. As for Orbited2, I see it as an evolution of
the current Orbited - in fact, it removes MorbidQ and adds WebSocket
support (when available). Apart from that, I assume it works just like
the old one - I'll be undergoing tests with it quite soon, and then
I'll post the results of a 'dry update' (no fixes or changes in code

No, the errors are not in the standard 0.8 as well. mcarter added some
bug fixes and a new version of the CSP (comet session protocol) to try
to shut some memory leaks. The idea was to test them and release as
soon as we saw they were stable.

Hi,

I think this is a good idea, even though the transition might take
some time. I am one of the few running on 0.8 but I am not sure if
it's worth stabilizing (which is almost done) or if it's better to
focus on back porting the fixes to Orbited 0.7.x.

What do you guys think?

Also I'm interested in cleaning up the orbited.org site, which needs

Sure, give me a week. I'm currently travelling and will do this once I
get back.

Hi everyone,
Orbited2 is a rewrite of orbited on top of eventlet; It uses websocket and
Comet session protocol as its transport layer. Orbited2 offers both tcp
socket and websocket proxying. So if you write a websocket server (using
node, eventlet, etc.) and you want it to work with existing browsers, just

Hi tejas,

I used your fix (in fact I had already added the document.domain stuff
all around the place where IE complained - but apart from that, all
changes you provided were just crucial and yet far beyond what I'd
ever try. I guess we're all pretty busy lately (I know I've been truly
absent) but it would it be possible to post a page on our home with

Hey Niklas,

I tried the URL you provided and the error was there as well. I
thought it would have to do with some configuration on the JS files
for a (non-)standard debug present. I would assume these errors do not
show up while using the current version - but I have only been to
0.7.10 so far - so what is the focus of those fixes, and would it be

Hi,

I'm trying an Orbited 0.8 with UTF-8 fixes (not by me) but am not
really making progress. I get this in the browser:

ERROR LOADING jsio/net/protocols/mspp.js : TypeError: Object #<an
Object> has no method 'getLogger' Orbited.js:136
ERROR LOADING Orbited.js : TypeError: Object #<an Object> has no

ok i see, as channels don't have to communicate with other channels, we can deploy more than one hookbox instances, each of them having his own set of channels.
it's really a good idea :)
sent from my super iphone

The scalability goals of hookbox are basically this: If you need 5,000
concurrent users, then don't worry about it; If you need 50,000 concurrent
users, then buy a top-end server and don't worry about; If you need 500,000
concurrent users, then shard your application such that you can keep certain
users/groups on specific hookbox instances.

Sounds very promising, congratulations Michael. I will try it by integrating
instant update in some parts of webpages of an existing django application I
did (not spool (which already have orbited+xmpp, another one). I'm wondering
how easy it is to integrate.
"Hookbox is built with eventlet which runs on top of epoll/kqueue/etc., so

So I dare guessing you're preparing a broker for HTML 5.

Have you considered pluggable functionality, say, in case I want to
use stomp to talk to Hookbox and I don't want to create a separate
layer/server in between? It would be nice if messages could be
internally forwarded to an interpreter and then re-formatted to the

Hookbox is much higher level than Orbited, and provides far better
integration with existing web technologies. Hookbox tends to be far more
opinionated than Orbited. You can't use your own message queue, or your own
protocol: You do it hookbox's way or choose something else.
Hookbox is built with eventlet which runs on top of epoll/kqueue/etc., so

Hi Leon,
Hookbox is a message broker itself. In the future we may allow you to
replicate to amqp or stomp brokers. At the moment the project is geared
applications which can be sharded to about 10k users (or less) per instance.
-Michael Carter

Very interesting concept.. How does this compare to Orbited and is it
a viable/stable alternative?
How about scalability and performance?

Thanks..and great job!

--
M.

Michael,

I took a look at the documentation and would like to ask whether you
plan on handling the message broker within Hookbox itself. I am
currently using Orbited and on the production server I went with
ActiveMQ. Are you planning on developing a robust embedded message
broker that will be internal to Hookbox?

Greetings,
As some of you may know, I've been working on an alternate
comet/websocket/web-push server called Hookbox ([link]) for the
past year. The goal of the project is to expose much higher-level apis than
Orbited, and provide tight integration with existing web technologies, like
PHP, Django, Rails, Java Servlets, etc.

None of the Trac section pages work btw... Please advise. Thanks.

Hi John,
On Wed, Jul 14, 2010 at 7:07 PM, John Porter Simons
Thanks for letting us know. I'll have a look now.
Regards,
Matthew

I logged in with bug/html and went to [link], but
that page is down...

Also when I hit o.helloexpert.com:8000/static i do find my files so Im
not sure if my proxy is not working because nslookup o.example.com
points to my ip. Is there something basic that I am missing?

Hi, Im trying to setup a reverse proxy for my server so that it allows
people behind firewalls to use our system. But when my application
loads it unable to fetch the orbited.js files

My setup is as follows

apache.conf

<VirtualHost *:80>
ServerName example.com
DocumentRoot /var/www/trunk/public

@BobDobbs Thanks that might help if I decide to switch back to Rabbit.
For now..it seems like everything is working as expected..except for
the exception that gets thrown away once in a while..
As far as talkrespond.dev..that's something I added manually (wasn't
part of my code as in the code it's a variable)..I just forgot to

@leon.. yeah it's added to my /etc/hosts..I m not piping any traffic
to the webserver though...

Hey.. just reread over the post and realized why you got the localhost
auth issue. You were using the demo and the demo expects to connect on
localhost (its hardcoded). Try editing the demo javascript to connect
instead to your message broker. Then you should have js that is
expected to work and could maybe eliminate some variables. Sorry for

How's the setup for talkrespond.dev? Is that added to your /etc/hosts
file? Are you piping the traffic through your webserver (Apache?)?

In the past when I have seen lines like these:

Unauthorized connect from
'192.168.133.1':55569 to 'localhost':61613

It was related to how localhost is resolved. Sometimes host files are
weird or orbited doesn't like the name you gave it. I would generally
try opening orbited and editing the lines like this:

ActiveMQ was a breath of fresh air..not sure about performance, but it
works out of the box (since it's Java) and has a sweet web based admin
console.
Anyways, I was able to get my stomp client to communicate with
ActiveMQ through Orbited just fine... however every once in a while I
would get this exception on the console: (after which the connection

When I first tried, I went directly with ActiveMQ. I remember
considering RabbitMQ, but then the Apache background made me choose
ActiveMQ. It took me a while to figure how to set up the
authentication as I'm a complete moron in Java, but I described the
simple authentication plugin usage on a page you can find in this very

Yeah.. I am able to access the /static/demos url just fine..although
in the STOMP test demo I am unable to connect properly....rabbitMQ
allows me to publish messages with the guest account just fine though.

This is what I get in the STOMP demo after I try to connect:
→ Transport closed (code: 106)

After you start Orbited, are you able to visit the examples that ship
with Orbited? They should be available on [link].
If the page pops up, first step is fine.

I'm not into RabbitMQ, so I am not sure there's something else in its
conf file that could give you those problems.