Just a short note,
Now that our web app is coming together we were able to give our Darkstar servers a bit of a work out last night Results, though still preliminary, were very positive. With about 12 people going in and out of game sessions with 3 to 4 players per session everything worked, nothing died, and we saw almost no Darkstar server load.
With any luck we should be able to move to testing an order of magnitude more players in the next week.
Meanwhile, some thing we've learned about Web2.0 along the way.
(1) The Netbeans 6.5 profiler is totally broken with regards to Glassfish 2.0. Really, people, this is NOT good. The current release of two products central to Sun's Web 2.0 strategy don't work together. Even worse, there is silence on this fact such that we wasted a week trying to figure out what we were doing wrong before we gave up and went to a commerical profiling solution. And for any Sun folks reading, a quick translation: a week in a start up is about the equivalent of 2 months of development Sun time.
Sun is my alma-mater and I love it, but I gotta be frank: this is NOT the way to win customer loyalty to your software products. The NB profiler is really about the only thing NB has going for it when compared to Eclipse. Breaking it is a bad bad thing.
(2) Bayeux is a bad joke. Its a messaging protocol designed by people who know nothing about building secure online systems. Its biggest weakness is that it is ONLY a client to client protocol. A clue for any Bayeux people listening... that makes it almost worthless in the online game space, chat space (where you want to do server side message filtering) or almost any other at all sophisticated web app. We're struggling with it now but we will be replacing it soon with a custom built, true Comet based message hub that allows us to push messages from the server as well as from clients and in general have the server involve itself intimately in the message stream..
Other then that, our hodgepodge of web technologies is working out fairly well. I'll be submitting a talk to AGC where I will dissect what we've built, what we used, and all the lessons learned. I submitted that talk to CGDC which is earlier... and they turned it down. Which frankly astounds me. BUT of course the blindness of CGDC to the online game phenomenon is part of what made AGC happen to begin with so maybe I shouldn't be so surprised.
Now that our web app is coming together we were able to give our Darkstar servers a bit of a work out last night Results, though still preliminary, were very positive. With about 12 people going in and out of game sessions with 3 to 4 players per session everything worked, nothing died, and we saw almost no Darkstar server load.
With any luck we should be able to move to testing an order of magnitude more players in the next week.
Meanwhile, some thing we've learned about Web2.0 along the way.
(1) The Netbeans 6.5 profiler is totally broken with regards to Glassfish 2.0. Really, people, this is NOT good. The current release of two products central to Sun's Web 2.0 strategy don't work together. Even worse, there is silence on this fact such that we wasted a week trying to figure out what we were doing wrong before we gave up and went to a commerical profiling solution. And for any Sun folks reading, a quick translation: a week in a start up is about the equivalent of 2 months of development Sun time.
Sun is my alma-mater and I love it, but I gotta be frank: this is NOT the way to win customer loyalty to your software products. The NB profiler is really about the only thing NB has going for it when compared to Eclipse. Breaking it is a bad bad thing.
(2) Bayeux is a bad joke. Its a messaging protocol designed by people who know nothing about building secure online systems. Its biggest weakness is that it is ONLY a client to client protocol. A clue for any Bayeux people listening... that makes it almost worthless in the online game space, chat space (where you want to do server side message filtering) or almost any other at all sophisticated web app. We're struggling with it now but we will be replacing it soon with a custom built, true Comet based message hub that allows us to push messages from the server as well as from clients and in general have the server involve itself intimately in the message stream..
Other then that, our hodgepodge of web technologies is working out fairly well. I'll be submitting a talk to AGC where I will dissect what we've built, what we used, and all the lessons learned. I submitted that talk to CGDC which is earlier... and they turned it down. Which frankly astounds me. BUT of course the blindness of CGDC to the online game phenomenon is part of what made AGC happen to begin with so maybe I shouldn't be so surprised.
1 comment:
Hi Jeff,
About the problems with the profiler of Netbeans: The best way that I found to stop to complain (too late) about the bugs was to join the netcat program (community driven QA phase) for netbeans 6.0 and 6.5. The result is that a lot of platform bugs that were bothering me 2 years ago are gone.
I think that the profiler has not been tested a lot by the netcat participants because it may have not been their focus and/or center of interest.
You are welcome to join the netcat and enlarge the diversity of the user pool ... and in the end, you can get a nice Netbeans TShirt :-)
Regards,
Vincent Cantin (karmaGfa)
Post a Comment