Switching over to SSE: an informed decision.

[Gabriel Benmergui]

Hey. I know Alex deprecated Juggernaut already, but im still using it in my production servers.

This is a big architectural decision so im reviewing my application from that standpoint . Setting up the SSE server seems trivial, and switching all the juggernaut code to this protocol is quite trivial also.

My concern is how reliable SSE's really are compared to socket.io. Although SSE's are native (althought not supported on IE) i dont know the performance comparison of running sinatra+thin vs node.js socket.io.

Any tips on how should i approach making this decision?

[Alex MacCaw]

Yes, sorry, I don't have any numbers on this.

My guess would be much more reliable, since WebSockets have historically had problems with proxies etc modifying their packets.

The various shims for SSE basically turn it into long polling - a reliable transport.

Alex

--
You received this message because you are subscribed to the Google Groups "Juggernaut 2" group.
To post to this group, send email to juggernaut...@googlegroups.com.
To unsubscribe from this group, send email to juggernaut-for-r...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/juggernaut-for-rails?hl=en.

--

Alex MacCaw

+12147175129
@maccman

http://alexmaccaw.com

[Yves-Eric]

Hey Alex,

My first reaction when I saw Gabriel's post was "What, deprecated? He must be kidding!". Then after your reply: "WTF? I must have missed the memo...". Except, looking at the list archives, there was no memo...

Many rely on mailing lists to keep up to date on the status of the several gems and libraries they use. Could we be nice to them and have an official announcement here, as a new topic for best visibility?

Cheers,

PS: Great work on Juggernaut, it served us flawlessly in production since 2008. Glad to see it is going away for the best of reasons!

-- 

Yves-Eric

On Thursday, 30 August 2012 06:30:15 UTC+9, Alex MacCaw wrote:

Yes, sorry, I don't have any numbers on this.

My guess would be much more reliable, since WebSockets have historically had problems with proxies etc modifying their packets.

The various shims for SSE basically turn it into long polling - a reliable transport.

Alex

On Wed, Aug 29, 2012 at 10:01 PM, Gabriel Benmergui <gabriel....@kaya.gs> wrote:

Hey. I know Alex deprecated Juggernaut already, but im still using it in my production servers.

This is a big architectural decision so im reviewing my application from that standpoint . Setting up the SSE server seems trivial, and switching all the juggernaut code to this protocol is quite trivial also.

My concern is how reliable SSE's really are compared to socket.io. Although SSE's are native (althought not supported on IE) i dont know the performance comparison of running sinatra+thin vs node.js socket.io.

Any tips on how should i approach making this decision?

--

[Alex MacCaw]

Yes, I dropped the ball there. Will post an official announcement.

--

You received this message because you are subscribed to the Google Groups "Juggernaut 2" group.

To view this discussion on the web visit https://groups.google.com/d/msg/juggernaut-for-rails/-/GZ8ByJnUnvgJ.

To post to this group, send email to juggernaut...@googlegroups.com.
To unsubscribe from this group, send email to juggernaut-for-r...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/juggernaut-for-rails?hl=en.

[Gabriel Benmergui]

Thanks for the response. Currently i get a certain number of users unable to connect properly , and this is one of the reasons i am considering making the change.

Once i had to deal with port handling, becuase without port 80 many anti-viruses blocked it. However, some antiviruses still block the port, and even after that, i do get some people unable to connect. Even worse, it would seem they are able to connect(trigger the connect callback) but cant receive any data.

I dont know if its a socket problem, or juggernaut problem, but it hits me hard. I just wouldnt want to make an SSE switch and find out it has poorer performance, and similar issues. 

I guess i have no other option but to give it a try. Do you know of any site that has built itself with SSE properly? id like to run ntests on it. At least, ask my problem users to see if they can also access that service.

To unsubscribe from this group, send email to juggernaut-for-rails+unsub...@googlegroups.com.

For more options, visit this group at http://groups.google.com/group/juggernaut-for-rails?hl=en.

[Gabriel Benmergui]

Also, to note. Some people have made another socket.io protocol, SSE. It hasnt made master probably because its not what the socket.io guys have in mind.

So maybe it would be a good use case to try to configure socket.io for juggernaut to use SEE by default, or the usual fallbacks if unavailable and disable websockets altogether.

It certainly would be good for my use case, to try out SSE without having to do code changes on my clients and servers

. 

How does that sound?

What i noticed about using SSE is that in case messages are lost, i will have to worry about keeping track of them myself. (that is, using message ids, and stuff) and im not a fan of that.

Regards.

To view this discussion on the web visit https://groups.google.com/d/msg/juggernaut-for-rails/-/cUIo6NU9hx0J.

To post to this group, send email to juggernaut...@googlegroups.com.

To unsubscribe from this group, send email to juggernaut-for-r...@googlegroups.com.