Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[thegoatgod_pan]

The most worrisome part about this whole thing, is that there really isn't a word about the game itself, except in the vaguest of terms "it will be released!" "still working!" etc. It is as though it were completely dead in the water.

You'd think they'd rectify the situation with an update of some kind: an image, an inkling of what has happened over the last few months, anything...

I am also pretty sure (while not a lawyer) that you cannot decide not to refund 60,000$ of pre-purchases without getting sued, or pass the buck to paypal (again, without getting really, really sued) if the game never materializes

[Delta Foxtrot]

From the way they said it, it's not that they choose not to refund. It's not like they banked the entire KS pot for a rainy day.

[Sergarr]

The determination of the coder is one of those things that's been questioned.

I reckon it's less about "determination" and more about "general programming know-how"; his AI bug, for example, is a good example of bad programming practices. Recursive function calls in situations where existence of a stable solution is not guaranteed, and without any random component to smooth it out? It's one of those things that first-timer programmers tend to do, because it sure does "look beautiful" (I know because I remember doing something like that when I was new);

If he was more experienced and thus started to work from already existing architectures (of which there are many), which incidentally never use recursion, he at the very least would not have had that problem, which would've accelerated his game development by 1-2 weeks - and that's only on that one problem. He undoubtedly have had many more, of which he hasn't told us, and of them, there were probably many that could've been avoided, saving literally months of coding time.

[Xgamer4]

Apparently, from what I remember Sean saying, he managed to partly oversimplify and partly misrepresent the actual bug (not too surprising), and the actual problem was related, but nowhere near as basic as you're saying, and we were believing.

But I tried to dig up his comments, and decided it really wasn't worth my time... those update topics on the official forum are trainwrecks - nothing more. So take what I remember with a grain of salt.

[Serenseven]

I'm one of the moderators of the official forum, and also the individual who was asking questions of Josh each week and relaying the answers to the group before communication broke down completely.

The version of the AI issues that was presented here was an oversimplification of the already oversimplified version that I presented to the official forum. You have to remember that I was faced with the problem of how to make advanced programming issues intelligible to an audience who mostly have never touched a line of code in their lives, and also that I'm not a Unity programmer. So I'm trying to take a couple paragraphs of highly technical information about a programming environment with which I am not familiar, and explain it to complete novices. Necessarily, I simplified a bit, and went for more of a general analogy rather than trying to present a code-accurate version of the problem.

A backer took that explanation and posted a condensed version on the public section of the official forum, which in turn was abbreviated further when it got quoted here. By that point, it was in the form of a simple recursion problem any freshman computer science student should handle with ease. The original was (and maybe still is, for I'm not sure it's been 100% fixed) far more difficult, involving multiple simultaneously-called objects not handling their data properly when called through particular interfaces - in other words, some sort of thread safety issue. Multithreading with shared resources can be a pain in the arse even for experienced developers.

TL/DR: While there may be any number of valid reasons to question Josh's competence, determination, or whatever at this point, the AI issue - as presented here - is not one of them.

[Xgamer4]

Heh, I wander off to spend a few more minutes trying to find the actual post and the guy himself jumps in.

I did find the post I was thinking of, but it has a big, fat *DO NOT SHARE* title on it so... hopefully my one-sentence summary isn't too much sharing.

For those able to access the backer forums, it's in the Dec 8th topic, page 4. If you're not afraid to manually edit URLs, it's f=4, t=395, p=16376.

[Sergarr]

...I have no idea how can one "simplify" a multi-threading problem to a recursion one. They're not even close. Like, at all.

There's a good term for multi-threading problems, https://en.wikipedia.org/wiki/Race_condition.

Also I must question this statement:

TL/DR: While there may be any number of valid reasons to question Josh's competence, determination, or whatever at this point, the AI issue - as presented here - is not one of them.

, because it sounds like he did not properly separate AI from the main game engine itself, which gave rise to the thread conflicts and such.

Unless TWS AI somehow requires more than 1 core to process, he could've put all AI parts into a single thread separate from the main game engine, and make it either run immediately after the main game engine process, with AI output being utilized for next game engine tick, or concurrently, but with a delay in inputs (i.e. gameworld is processing tick 30, while AI output is being calculated from a frozen gameworld "image" at tick 29).

I shall note that this kind of architecture (game engine separate from AI) is suggested as significantly more preferable in Game AI literature sources that I've read, precisely because other people have run into these kinds of issues, too.

[jhxmt]

Still not fussed.  Still happy to wait.

We can theorise and worry and murmur about the putative AI issue all day long, but without knowing the details it's not something I'm going to get particularly concerned about.  I'm not interpreting any of the messages thus far as particularly negative.  Yeah, there's been a lack of positive communication.  I'm okay with that.  I can understand why others wouldn't be.  At the end of the day, I take the view that it's more important for the coder to be coding than for the coder to be letting people know that he's coding.  Again, I appreciate others' mileage may vary on that.

[Serenseven]

Unless TWS AI somehow requires more than 1 core to process, he could've put all AI parts into a single thread separate from the main game engine, and make it either run immediately after the main game engine process, with AI output being utilized for next game engine tick, or concurrently, but with a delay in inputs (i.e. gameworld is processing tick 30, while AI output is being calculated from a frozen gameworld "image" at tick 29).

I shall note that this kind of architecture (game engine separate from AI) is suggested as significantly more preferable in Game AI literature sources that I've read, precisely because other people have run into these kinds of issues, too.

There are multiple entities (several hundred to a thousand or more, depending on the map size) that have AI to process each turn, and they have their own dedicated thread pool that they run in separate from the UI threads and the engine threads. The AI threads actually run continuously; during the player's turn they evaluate entities that the player hasn't interacted with recently and precalculate those entities' responses for the next AI turn as an optimization to reduce the player's wait time between turns.

[sirvente]

I'm no programmer but that seems weird as hell, wouldn't that cause the ai to behave different based solely on how long you take to end your turn?

[Serenseven]

No. If you do something that affects one or more of the precalcuated AIs, it throws the precalculated response(s) out and ponders it/them anew.

[Cruxador]

I am also pretty sure (while not a lawyer) that you cannot decide not to refund 60,000$ of pre-purchases without getting sued

Legally, the kickstarter goals are not pre-purchases but donations, and there is no legal obligation to refund at all.

I'm no programmer but that seems weird as hell, wouldn't that cause the ai to behave different based solely on how long you take to end your turn?

The usual solution - that being the one used by Stardock, who pioneered this method in GalCiv2 - is to just make the time between turns take longer if it needs to.

[Neonivek]

No. Kickstarter isn't donation it is funding. There is a legal obligation to refund.

Kickstarter isn't a Charity (though certainly you CAN do charity through Kickstarter). You are asking people to "fund" your work. You have every obligation to ensure that they receive said product.

It isn't a pre-purchase either.

[Cruxador]

There is a legal obligation to refund.

Only if it can be proved that a good-faith effort at success was not made. Which isn't possible except in the most drastic outright scam cases.

[Darkmere]

No. Kickstarter isn't donation it is funding. There is a legal obligation to refund.

I need a source for this. All KS says is that developers should include refunds to people who have not received their reward, based on amount of cash remaining and the amount pledged. They're very careful about pointing out they have nothing at all to do with anything between backers and developers.