We might very well be interested in doing something along those lines for Tribal Trouble.
- elias
I think it’s already been mentioned that this is a “bug” that’s already being fixed for 1.5 ?
If someone can think of a fair way of doing this, then we could run the test on JGF with multiple games (whose authors were willing to take part). However, only if it’s going to be done scientifically, which means starting by stating a hypothesis and then describing how that will be confirmed or disproved.
It would be moronic to assume you could just put two URL’s side-by-side and get any meaningful results; a lot more thought is needed than that. Both in visual presentation (how do you measure the bias of people who don’t care; how do you measure the bias of people who see the word “windows” and click on it, not caring whether it’s an EXE or a JNLP? etc) and in technical implementation (how do you correctly measure the number of failed downloads that then succeeded later, e.g. after downloading the JRE, thereby not artificially inflating the “failure rate”? etc)
I also humbly suggest that E would have to be careful in their involvement with this, since it potentially makes it look like Excelsior is fighting a battle “against” webstart. Presumably your actual interest is in e.g. getting data to decide whether you need to add JWS-capabilities to JET exe’s.
I could gather some stats on this. It’s all in place ready to rock and roll although I haven’t recently managed to compile AF with Jet (lack of time)
Cas 
[quote]I think it’s already been mentioned that this is a “bug” that’s already being fixed for 1.5 ?
[/quote]
Ok, but do you know a gamer who has 1.5 installed but who is not a Java developer?
[quote]If someone can think of a fair way of doing this, then we could run the test on JGF with multiple games (whose authors were willing to take part). However, only if it’s going to be done scientifically, which means starting by stating a hypothesis and then describing how that will be confirmed or disproved.
It would be moronic to assume you could just put two URL’s side-by-side and get any meaningful results; a lot more thought is needed than that. Both in visual presentation (how do you measure the bias of people who don’t care; how do you measure the bias of people who see the word “windows” and click on it, not caring whether it’s an EXE or a JNLP? etc) and in technical implementation (how do you correctly measure the number of failed downloads that then succeeded later, e.g. after downloading the JRE, thereby not artificially inflating the “failure rate”? etc)
[/quote]
How about the following scenario:
-
A visitor comes to the game Web site and clicks the “Download” link.
-
If a certain cookie ZZZ is present and its value begins with “JNLP:” or “EXE:”, go to step 5.
-
The server tries to set the permanent cookie ZZZ with value “JNLP:” or “EXE:”, chosen randomly with 1:1 probability, concatenated with randomly generated visitor ID.
-
The server redirects to another script that checks if the cookie was set successfully. If not, it serves the combined download page and does not count this visit.
-
The server serves the respective download page depending on whether the cookie begins with “JNLP:” or “EXE:”, and amends the tracking log.
A download script can be used to track EXE downloads, but tracking JNLP downloads can be somewhat tricky given that e.g. AOL users connect through multiple proxies. At the very least, you could track download initiations.
The same cookie may be used to track sales.
Well, our actual interest is to (hopefully) show that good old EXE (not necessarily JET-generated) has advantage over Webstart in terms of your game download and sales figures. Do not get me wrong - Webstart is great for the enterprise environment, where all workstations are under your control, you have got system administrators to JNLP-enable all browsers and install the same JRE on workstations, network is fast, users can be trained, and so on. But as a consumer willing to try your game, I do not care if your game was written in Java, C, or FORTRAN. I just want to get it to my desktop with as little effort and bandwidth as possible.
This issue is not limited to Java - I (as a person) do not have the .NET framework on my computer, and I will not download any software package that requires it or includes it.
the reality of stats : nobody use the .jar on .jnlp. Actualy a lot of GAMER compuers don’t have the JDK1.4.2 you need some got 1.3, some got 1.4.1, some got MSJVM).
All the gamers prefer the 5MB .exe than the 20MB useless JRE …
-
The reality is that “gamers” will download what they need to play a game they want to play.
-
The reality is that “gamers” won’t, they’ll just give up.
-
The reality is that “gamers” would rather get everything on CD.
-
The reality is that “gamers” would rather write everything themselfs.
Actually, the reality is that gamers fall into rather alot of categories, hence the opening for lots of different types of games. Any statistics you draw would be open to interpretation in one way or another.
What you want to see is use cases. Unfortunately, this puts web start at bottom pegging since there arn’t any games that use it and have been commercial successful. Obviously there are a fair few that have been using a .exe.
However, it wasn’t that long ago that no 3D game had been commerically successful. Or any game for that matter…
Just offer both if it doesn’t hurt you too much.
Kev
PS. Also, the 5MB limit applies for a small subset of games, not for every case.
PPS. I still don’t really understand JET, why can’t I compile using AWT/Swing ?
You can, except that it immediately drags in another 10mb of code you have to distribute so you would probably be better off with the JRE (except for startup time)
Cas
Ah right, see your point. Did you happen to catch that Live Chat? Some indication that they were actually considering the stripped down JRE idea…
Kev
[quote]We might very well be interested in doing something along those lines for Tribal Trouble.
[/quote]
Email me at dleskov at excelsior dash usa dot com.
[quote]* The reality is that “gamers” will download what they need to play a game they want to play.
[/quote]
I think there will be many such gamers only if your game is a true hit (a sequel to a true hit), you are known for previously making true hits, or you run a huge advertisng campaign. Imagine Doom IV will need JRE 1.8.2_11 - the Sun Java download server will think it is a DoS attack. Otherwise, there are so many shareware games out there to try…
[quote]* The reality is that “gamers” won’t, they’ll just give up.
[/quote]
This is my experience. In a general case, if I download a piece of software (not necessarily a game) to address some problem and it does not install or run on first attempt, I usually delete it and forget it, unless (i) I used the previous version and was satisfied with it; (ii) I could not find an alternative or (iii) I have learned from somebody it will solve my problem completely, so that I have motivation for trying circumvent the problem.
[quote]* The reality is that “gamers” would rather get everything on CD.
[/quote]
For a large game this is probably true. But if you have 100MB of textures, adding the JRE does not make any difference anyway.
[quote]* The reality is that “gamers” would rather write everything themselfs.
[/quote]
[quote]Actually, the reality is that gamers fall into rather alot of categories, hence the opening for lots of different types of games. Any statistics you draw would be open to interpretation in one way or another.
[/quote]
Ok, let’s say we are talking about shareware games which demo versions, if written in C++, would not have exceed 5MB in size.
[quote]What you want to see is use cases. Unfortunately, this puts web start at bottom pegging since there arn’t any games that use it and have been commercial successful. Obviously there are a fair few that have been using a .exe.
[/quote]
Alien Flux is WebStartable.
[quote]However, it wasn’t that long ago that no 3D game had been commerically successful. Or any game for that matter…
[/quote]
3D makes a difference in gaming experience, whereas Java makes a difference in game development experience. So your analogy is not correct.
[quote]Just offer both if it doesn’t hurt you too much.
[/quote]
Right.
[quote]PS. Also, the 5MB limit applies for a small subset of games, not for every case.
[/quote]
See above. Also, I do not think the subset you are talking about is that small.
[quote]PPS. I still don’t really understand JET, why can’t I compile using AWT/Swing ?
[/quote]
You can, but you would have to include the JRE with your product. Click here for the explanation.
I remeber Cas saying nobody was using his .jar/.jnlp release of AlienFlux (according to stats) you can confirm Cas ?
The reality is, that a gamer will download a 10mb video if someone in the forum wrote “olol” next to it’s link.
Anyway, all this sounds fine, but is still pure theory.
Not really. There are already alot people (hardcore gamer category), who have a current JRE installed. Most of em installed it for a specific application (Azureus or rather odd things like a Go client) or because their browser came without Java and they wanted to play stupid applet games.
I think a survey would be nice to get some solid numbers.
The acceptance for webstart is pretty high, so far everyone prefers it over applet and the common installer procedure (I talked about it with approx 30 people [all non devs and casual to hardcore gamers]).
Don’t get me wrong. Jet is really a very nice product, but I’ll try not using it because it’s a bit against my logic (accumulating overhead with each application).
From Alien Flux post-mortem article ( http://puppygames.net/articles/alienflux_postmortem.php )
- Deployment with Jet
We suspected that the requirement to have Java 1.4 installed on the client machines would make our game an instant flop. Looking at the webstats we have now, we can see that we were right - we’ve had nearly twenty times as many downloads of the Win32 exe we built with Jet than the Java version. Nearly all of the downloads of the Java version were from Linux users.
It’s evened out a little but last time I counted, Webstart made up a paltry 10% of installs.
Cas
[quote]It’s evened out a little but last time I counted, Webstart made up a paltry 10% of installs.
Cas
[/quote]
But anyone reading this has to go check the PG website and then it becomes obvious some of the factors influencing this.
For instance, if you click on the download link, you get a page whose first button is “Alienflux for windows” “Download Now”. What kind of user who has windows is even going to bother reading the next paragraph?
If you re-arranged the page so that the first item was “Alienflux for Windows, OS X, and linux”, then of COURSE you’d have a heck of a lot more JWS users.
As it is, the page currently assumes that the player knows (or even cares) that they have this thing called “Java Webstart”. How many people are NOT going to get confused by that and think “but my OS is called Windows, so obviously that link won’t work for me”?
Obviously, Cas doesn’t care about trying to increase his number of JWS users so it’s not like he cares about the above; my point is merely that if you want to do a test on this kind of thing, in order to get useful stats you absolutely have to be very careful about the download page.
PS I’m surprised that the JWS figure is as high as 10%. That suggests to me that either you have almost 10% using linux, or else there are a lot more people around who know/care what JWS is then I would have guessed.
OR…there’s the one which IIRC is more than all those others, statistically: (iv) My friend told me it was worth suffering to get this game
alternatively: (iv-b) My friends are playing this (multiplayer) game and nagging me to download it so we can play together
[incidentally, as a master of Quake, and a die-hard Q3 fan, that was the only reason I ever got into UT long enough to get hooked: peer-pressure!]
Of course, that reason is currently almost entirely responsible for making Sony somewhere in excess fof $15 million a month (and others are making even more). “My friends are playing it” is (alleged by Sony’s dev team, and quite a few others, and some psychology reports on social trends) the biggest driver of all.
Strictly speaking, no. Webstart can make a considerable difference to the gaming experience (e.g. one facet: I don’t have to find any 3rd party libraries, download them, install them, etc. Number one reason I don’t play linux games: 99% of the time you have to go find obscure versions of 5 other obscure libraries that are probably mutually incompatible with later versions you already have installed)
I remember the DOS days of gaming, and how I became an expert computer user: it was because getting games to work was so goddamned difficult (I soon understood PC’s better than my parents, who were both programmers!). To me, JWS (and anything in the same league) is the next step forward on this continuing ladder from DOS-doldrums, through WiSE installer/installshield, to someday games which just work ;D.
Well, 1.5 isn’t out yet, so… And once it is, I’d imagine 60%+ of all windows 1.4.x users will upgrade, seeing as it’s now semi-automatic for them.
Sounds a good start. What about the percentage of “succesful starts” though?
I’m going to remove the Windows version for a bit and we’ll see how many installs I get with just Webstart. I’ll put those gubbins back on the download page that invite the user to install Java as well.
Cas
Since alot of people don’t know what webstart actually is… I would change that headline. “Alien Flux for Java Webstart” -> “Alien Flux for Windows and Linux (Java Webstart)” and the description text should point out, that you need Java. There could also be a small webstart app for checking if everything works (java+opengl).
I know very little about webstart, but after the experience I’ve had with it I’m not a fan.
If I download a webstart app, it just gets plonked away somewhere down the documents and settings directory route, and there’s no way to remove the app unless I trawl through loads of folders and manually remove the app myself. On top of that, once the game is installed I can’t play it off-line, I have to connect to the net every time I want to start it, its shit, well, version 1.42_1 is anyway.
What I don’t understand is that if you need Java installed on your system to use webstart then why not just link to a downloadable jar?
Are people that reluctant to download a jar file?
Btw, as many people have pointed out, I don’t think that downloading the latest JVM is that big a problem, if your game is desirable enough, etc.