Lots of doors are being closed for Java

General Discussions
#72

It’s almost shocking to think that applets first appeared almost 16 years ago. In time Java has had tonnes of development, but applets themselves have had almost none. They still feel very clunky and heavy when starting up, can be difficult for new developers to get working and have lots of odd bugs. On FunOrb they suggest minimizing and remaximizing if your applet fails to appear, and it’s not uncommon for other sites to write “if it doesn’t work then try refreshing the page”. You don’t get that type of flakyness with Flash or HTML+JS (at least not to the same extent).

But I think the main issue is that although applets are more then capable of out performing their rivals, it’s really difficult to do so. For example there are editors for building pretty UI’s specifically for Flash and Silverlight, whilst none for Java applets (as a result applets always tend to look like applets). Plus so many basic things have been missing like video support. Silverlight is only 4 years old and has none of the big issues that applets do (like fast startup, fully hardware accelerated graphics and video support).

#73

JavaFX was a spectacular waste of resources, the whole java client side team were mending that sinking ship for months with no result. There is still almost nothing to show for it years afterwards, and it promised video playback, fast start up, integration with adobe tools, integration with swing, jogl and the kitchen sink

For me one of the biggest boosts for java games was Ken Russell the engineer who made plugin 2 (and jogl), as well as Chris Campbell and Dmitri Trembovetski who were responsible for giving java 2d proper d3d performance on windows in 6u10.

#74

haha, I’m aware of that issue, its something LWJGL Applets also suffered from in the past but was fixed years ago, too bad funorb use their own opengl java binding :).

#75

Yup they could have easily been where Flash is now but Sun just didn’t care about that market.

#76

Fools in charge that’s why.
Actually JavaFX2.0 showed serious promise. And it’s still twitching. Might be something in it yet.

Cas :slight_smile:

#77

The reason I think Erlang is an unreasonable choice isn’t because of any technical issues, it is because of risk avoidance. Scaling shared state concurrency is a near intractable nightmare and Erlang’s model (actors/messages) is the only reasonable solution that I’m aware of. The problem is that Erlang is a highly complex functional language that one would have trouble finding a pool programmer capable of properly using it. As an aside, I find it interesting that Blizzard/Activision bought DemonWare back in 2007 and WoW nor any other Activision game (to my knowledge) uses Erlang. My thinking is that you can get the same set of features from other languages with less risk.

WRT: “It can handle a high number of threads” - We’re talking about two different animals here. I was talking about “real” threads at the OS/hardware level, whereas actor/message (et al.) systems use “lightweight threads” (fibers, picothreads, etc).

MMO Server (split)
#78

@Roquen: Exactly, and that’s why I said that serious, high-performance server-side programming is done almost entirely on the JVM. Akka is probably a great substitute for Erlang, and since more and more is moving server-side, I believe Java’s place in game programming will only grow.

@Cas: I have high hopes for JavaFX 2, too, but in retrospect, I’m not sure Sun’s negligence turned out so bad. The fact is that client-side is fragmented, and the only technology in consensus is the committee designed and non-proprietary HTML+Javascript. That’s the way the leading hardware vendors want it. They don’t want WORA. Apple doesn’t want Flash. The Playstation and the Wii don’t want to share games. Differentiated software sells differentiated hardware. No proprietary standard stands a chance. However, the client world is becoming so fragmented, with so many different types of devices, OSs and programming language, that developers simply won’t put up with it for long, and there will probably be one or two victorious software technologies. Should Oracle put itself into this dangerous competition, especially without client hardware of its own? Well, I’m sure people in Oracle are pulling their hairs out in frustration over Sun not inventing Android, but that ship has sailed. What should they do now? It’s a very difficult question. The best result they’re hoping for (and the one which will probably be best for the Java community) is some sort of a favorable settlement with Google over Android, which will allow them, and “standard” Java to put its foot in that door.
Other than that - they should hold on and fortify their position on the server, and do well with JavaFX2 to make up for lost developer mindshare.

#79

It can be elegant to have a thread per connection, and because of Erlangs green threads it makes applying this model more straight forward then with say Java. For example there are many comparisons between various Erlang built web servers and Apache, where the Erlang solution can handle a much higher number numbers of connections before falling over and during that time the amount of data it’s outputting barely drops (unlike Apache). Secondly Facebook uses Erlang for their messaging system specifically because it means they can have millions of threads all waiting for data to come in. Using greenthreads is ideal as most of the time the vast majority of those threads are blocked. In short you don’t have to worry if your number of threads spawns into the millions where as with Java and others you need to apply a solution to deal with that (thread pools, runnable tasks, etc).

I learnt Erlang for my final year project at uni and tbh it took me about a week or two to be able to start building complex system with it. It’s primarily just the pattern matching and single assignment that makes Erlang a functional langauge; it’s really on the edge of the definition and it’s much simpler to learn then proper functional languages like Haskell.

MMO Server (split)
#80

That advice on FunOrb predates Jagex using OpenGL. I think the biggest problem was caching the Graphics object and using active rendering.

#81

And they are fast on 80’s hardware aka slow cell phones. Do you expect Quake 3 on a 386? My games are much higher performing than yours on BlackBerry, J2ME, and Android phones.

#82

I think you just made an apples to luxury yacht comparison there…?

Cas :slight_smile:

#83

Nothing to add :slight_smile:

#84

You should send dog poo in the post to them, Chris.

Cas :slight_smile:

#85

Mono Touch requires Mac OS X (Xcode) and Mono .NET is not really cross-platform, it is noticeably different than Microsoft .NET implementation (look at my identi.ca account, I have given some examples about it), it is not comparable to Java. Lol DirectX is not really useful, OpenGL is better supported especially on embedded platforms, Direct3D Mobile is not very famous.

Do you know Myriad Alien Dalvik?
http://www.myriadgroup.com/Media-Centre/News/Myriad-Announces-Alien%20Dalvik-Enables-Android-Apps-to-Run-on-Non-Android-Phones.aspx
Android applications can now run on non-Android platforms.

It is not my fault.

I agree with you. We cannot do miracles on mobile phones that have a pretty bad implementation of J2ME.

#86

From my point of view the problem with using Erlang for a game server isn’t the language or the concurrency model, the problem is: are you going to code all the game logic in Erlang too, or do you want to use scripting? What about physics and collision detection (which has to happen on the server to some degree to avoid cheating)? Code all that in Erlang or use pre-existing solutions?

Beyond the mostly trivial (at this point using libraries similar to enet) client server communication, there’s a whole lot more going on in the server. Erlang most likely has ways to integrate with scripting languages but how well supported are they? Implement your own? Check Erlangs string handling ability before setting sail in that ship. Collision detection and even rudimentary physics? Erlang strenght is not speed in numerical computations.

So yeah, there’s a lot more to consider than the concurrency model.

MMO Server (split)
#87

That’s a really good point.

#88

This is such an interesting thread I just thought I’d throw in my 2c worth of rant…

My background is in academics (I teach first and second year computer science). Java is still the dominate language here, mostly because of external pressures during the early 2000. As I have an interest in gaming and I was forced to use Java, I decided to create a simple 3D gaming library in Java for my students (using lwjgl for rendering). Since then, the library has evolved into something rather useful and I am interested in the Java client-side deployment options.

During my research, it seems that Java is indeed very popular in the big shops on the server side, but almost non-existent on the client side until Android comes along.

This is very discouraging and frustrating at times, as I keep running into walls when I tried to create applets that run reliably on all platforms, windows seems to run fine, os x is a pain (their own Java version does not play well with applets), linux is ok but not great. When apple come out with their App store for macs, they outright reject all Java apps, making distribution of Java games on the mac desktop impossible. Android is the new hope for large scale Java deployment but Oracle is now suing Google over it, further hampering the potential of Java.

It seems to me that Oracle should invest into Java technologies (webstart, applet, etc.) instead of litigation.

#89

I would agree with the fact that Oracle MUST improve deployment for client side ( but… but… I must admit that I am really happy that google is suing by someone that have enought power to sue it )

#90

Don’t rely only on applets (especially on Mac, even the Ardor3D team has had some problems with it) and personally my first person shooter (that uses OpenGL but not LWJGL) works quite fine on Linux, Solaris, Mac OS X and Windows, someone succeeded in installing it on AmigaOS (I don’t know how).

I would like them to work together because Android is now a big player but the DVM has worse performances than J2SE For Embedded.

#91

FWIW I’m revisiting XMLVM after a year or so and it’s improved greatly. Performance isn’t terrible any more and much more of the JDK is available.

I’m going to sort out some performance tests if any one has any ideas for a suitable targets?

Kev