Hey, I'm new here and I have a question - more out of interest than that I'm likely to do anything with it.

To introduce - I had some mates round this weekend and while they're not into free/libre games regularly, we all played some Nexuiz and had a blast . So good job for everybody involved!

I've also been playing Tremulous for a while now, and so I couldn't help but wonder that you guys are using two different engines (though of course diversity has advantages as well).

I've had a bit of a look around, and I read that people reckon Nexuiz would look worse on the ioquake3 engine then on Darkplaces - is that still the case? Are there other reasons that speak against porting? And what reasons might Tremulous have for not switching to Darkplaces?

Thanks for your time,
a very curious Sean

First of all: switching to another engine means throwing everything away, and rewriting it from scratch. You really only do that if you have to.

Technical advantages of DP over ioq3 that Nexuiz really should not lose:
- more advanced rendering features, eye candy
- easily customizable menu
- more intuitive command line interface
- HTTP downloads for missing maps (nexuiz map packages often are HUGE)

Technical advantages of ioq3 over DP that Tremulous really should not lose:
- much less CPU load on the dedicated server, as no interpreter is running but native code only
- AAS support for bot AI
- much better handling of player prediction

The point of the original decision was, indeed, the advanced rendering features.

It seems a little late in the game for this question, since imo, the big question now is how will things compare when id releases its 4th generation engine?

Flying Steel wrote:It seems a little late in the game for this question, since imo, the big question now is how will things compare when id releases its 4th generation engine?

DP already has most if not all of the rendering features of id Tech4, except for maybe Megatexture.

Exactly.

Main things missing is good content. The engine itself is quite up to date and comparable with id tech 4, while ioq3 is somewhere between id tech 3 and id tech 4, in terms of visual quality.

But, our player models look really bad, and many maps have blurred textures. THAT is where we currently are lacking. No engine change can fix THAT.

I hear you on the content issues, I can help on that front.

I was thinking of non-graphical modern features though, like vehicle physics. I wonder what id tech 4 could bring to the table there.

The engine has not much to do with that.

In fact, vehicle physics are in Nexuiz svn, but at the moment only one vehicle.

The question remains however if vehicles would even be of any use in a fast paced game like Nexuiz.

divVerent wrote:The engine has not much to do with that.

In fact, vehicle physics are in Nexuiz svn, but at the moment only one vehicle.

The question remains however if vehicles would even be of any use in a fast paced game like Nexuiz.

And what about tris limits? How many tris can I "fill" in a map to avoid slow motions on , let's say, a dual core modern pc?Another question: I saw on http://svn.icculus.org/twilight/trunk/phystest/ a piece of code about ragdoll, I compiled it and.. it's amazing!Is it just a test or one day we could see players swept away from a shotgun blast?

divVerent wrote:The engine has not much to do with that.

In fact, vehicle physics are in Nexuiz svn, but at the moment only one vehicle.

The question remains however if vehicles would even be of any use in a fast paced game like Nexuiz.

I find vehicles useful and fun in nexuiz.

divVerent wrote:In fact, vehicle physics are in Nexuiz svn, but at the moment only one vehicle.

Do you think we can get this vehicle in the game by default for the next release? Like one or two placed in Reborn or Silver City somewhere?

Just to help the testing process of course.

The question remains however if vehicles would even be of any use in a fast paced game like Nexuiz.

I am sure they certainly will, it'd just be a faster vehicle with more firepower or it would be useful on bigger maps.

Character class mutator(s) also provide an opportunity to lower the game speed for tactical gameplay, by making all but the fastest classes as quick as default Nexuiz player physics. A much better solution than havoc was, imho.

toneddu2000 wrote:

divVerent wrote:The engine has not much to do with that.

In fact, vehicle physics are in Nexuiz svn, but at the moment only one vehicle.

The question remains however if vehicles would even be of any use in a fast paced game like Nexuiz.

And what about tris limits? How many tris can I "fill" in a map to avoid slow motions on , let's say, a dual core modern pc?Another question: I saw on http://svn.icculus.org/twilight/trunk/phystest/ a piece of code about ragdoll, I compiled it and.. it's amazing!Is it just a test or one day we could see players swept away from a shotgun blast?

that's old code that was a test for the original dpmod and alien infestation mod. lordhavoc just stoped working on it but now is working on twig.

Mizu Kitsune wrote:that's old code that was a test for the original dpmod and alien infestation mod. lordhavoc just stoped working on it but now is working on twig.

Thanks Mizu Kitsune!! I hope coders can embed something similar in the future . I took a look at http://urre.quakedev.com/ and my jaw felt on table!! This would be great in conjuction to vehicles and ragdoll!

np. cant wait to see the finished twig or even what is current, integrated into nexuiz/zymotic.

Sorry for the late reply - but many thanks to the plenty insightful answers!

I'd not thought about id Tech 4 being release - and I guess you're right, it's probably just a matter of time.
Interesting as well to here you think Darkplaces (and ioquake3 to a lesser extent) are pretty much on par with the commercial engines!

Two things were mentioned I didn't quite get though - one regarding the ioquake advantages for Trem: dedicated server and Bots I don't see as very relevant (there are no Bots I know of, and dedicated server load, well...). Player prediction though - does that have to do with showing players to one another, and therefore aiming ? (I read a discussion on the Trem forum about people leading their shots, but it perhaps no longer being necessary with a certain mode enabled).

The other thing is on porting - I had assumed that the greatest amount of effort in Nexuiz was the models/ animations and levels (what I would have regarded as content). I had kindof assumed they would still be usable. But I guess the graphical weapon effects and such are not portable. Is that what you referred to when saying it would mean "rewriting stuff from scratch" ? (For Trem I would guess that a good bit of work is in making the gameplay work, with buildings, evolutions, buying equipment and such).

Of course, if it looks prettier with the current engine, then I fully agree with you there's no point in switching.

(My original thought behind asking had been whether a combined effort might go towards making a better product then either effort separately. I'm a regular player of some Spring engine games - that is RTSs - and I often ponder how we ended with about 3 different engines in that segment...)

P.S.
Thumbs up if vehicles were included in future versions, by the way !

Edit: P.P.S
Trem apparenty has http download (since only very recently) . Can't say if they did that just for themselves, or whether its ioquake3 side (or even whether I got mixed up and misunderstood things :/...)

SeanHeron wrote:Sorry for the late reply - but many thanks to the plenty insightful answers!

I'd not thought about id Tech 4 being release - and I guess you're right, it's probably just a matter of time.
Interesting as well to here you think Darkplaces (and ioquake3 to a lesser extent) are pretty much on par with the commercial engines!

Two things were mentioned I didn't quite get though - one regarding the ioquake advantages for Trem: dedicated server and Bots I don't see as very relevant (there are no Bots I know of, and dedicated server load, well...). Player prediction though - does that have to do with showing players to one another, and therefore aiming ? (I read a discussion on the Trem forum about people leading their shots, but it perhaps no longer being necessary with a certain mode enabled).

No, this is about predicting your OWN location in case of lag.

The other thing is on porting - I had assumed that the greatest amount of effort in Nexuiz was the models/ animations and levels (what I would have regarded as content).
[...] (For Trem I would guess that a good bit of work is in making the gameplay work, with buildings, evolutions, buying equipment and such).

No, converting data files is usually quite easy, especially as very similar formats are used between Q3A and Nexuiz. The real work lies in all the CODE. It would take over a year to rewrite all that in another engine. Sorry, but this comment quite insulted me, as you apparently think all my work is worthless and anyone could do it in some weeks.

Making the gameplay work is actually the highest amount of work in Nexuiz development, followed by models, and only then the maps. Especially the support for many game types, and many map entities to allow maps to be more dynamic, and the bot code, are huge amounts of code. The menu system also was very much work, and really is much more advanced than Q3A's. You can ask the Urban Terror guys how much work their menu for the Q3A engine was - they'll tell you the same

If you like numbers: according to sloccount, an industry standard "retroactive cost estimate":

The engine has 117737 total source lines. According to the COCOMO model, developing it would take 15 developers 2 years. Given that half of the design and about 10% of code is by id software, make this about 1.5 years for 12 devs.

Game code has 70748 total source lines. According to the COCOMO model, developing it would take 11 developers 1.6 years. As all the design and code is entirely by Alientrap (nothing in there is from Quake, as Quake is MUCH simpler), this model should actually fit.

And in the last case, the numbers actually fit - 17.5 person-years can very well be right, given that Nexuiz has been developed for over five years now, although with a lower number of developers.

Sorry, certainly no offense meant :/ ! I was just stating what I as an unknowing user had assumed - I really don't have much of a clue to be honest.
I must admit that when writing I was thinking "hmm, all the different game modes" - and I'd completely forgotten Bots. And the Menu of course is kick ass, and makes a lot of the polish of the game of course! I guess I kinda glossed over that as it's not something you use while your actually fragging... .
In any case, I'll quite readily believe that there is a large amount of effort put into the code - I simply didn't realise (or was just coming to realise) that when I wrote the last post.

In no way did I mean to diminish your work! (Didn't realise alot of it was yours in fact...), please assume ignorance, not malice!

--
Thanks for clarification on prediction (though I didn't quite get it - it doesn't really matter).

I guess the conclusive answer to my general question then is that there's not going to be a "unification" of the open source FPS engines anytime soon (which I guess is just as valid as with the RTS engines - don't see it happening there either). A shame in some ways I guess, but there is certainly value to diversity as well!

Speaking of idtech 4, I wonder if the shadow code will also be included, or will it be removed from the sources? As many know, "Carmack's Reverse" was patented by Creative Labs in 1999(which is a joke, IMO). Even though Carmack independently discovered the algorithm on his own, id had to work out a deal with CL in order to retain it's shadow code.

LordHavoc, as well as I have done in CRX have been able to avoid the patent(at least in our estimations) by changing the algorithm sufficiently enough.

So I wonder, will the release of this code finally end the ridiculous patent by CL, or will the code be altered/removed?

I personally LOVE Darkplaces and ioquake3 a lot, and I've got Nexuiz (uses Darkplaces) and OpenArena (uses ioquake3) on my laptop (it's got a real shitty graphics card: an Intel 965GMA Chipset, BTW) I've tried both games and they're both great. But I've noticed that the engines' focus on completely different functionalities:

Darkplaces:
Naturally, DP focuses on graphics: advanced rendering, extreme OpenGL-driven effects, etc. The ioquake3 engine will probably NEVER be able to support these cool effects that we all know and love on DP. But its eyecandy-oriented design heavily impacts the FPS in-game. This impact is most visible on lower-end machines (such as my lappy, which typically gets around 12-30 FPS in Nexuiz). Also, the physics are a bit less realistic that iq3.

ioquake3:
This engine is just the opposite of Darkplaces, in a way. It's designed so that it can give you decent and not-too-impressive graphical effects, and replaces it with extremely high FPS. For example, whenever I run DP with everything set to its lowest setting, except for Realtime Dynamic Lighting, I get 9-26 FPS on my shitty i965GMA. But with ioquake3 I can enable RDL and get 75-95 FPS (and with RDL off, I can get over 100 FPS! Amazing, huh?) It's gives you Damn Small Nexuiz-type speed, but with OK-ish graphics and textures. Unfortunately, if you've experienced DP's graphics before, you'll probably get sick of iq3's low-medium graphics eventually and immediately start up Nexuiz (OA has rather low-poly weapon models, no reflections, HDR, gloss mapping, or decals and particles cannot be customized).

I guess that you can't have awesome graphics and consistent FPS on lower-end hardware. I like DP and iq3 both. Which do you like?

Sean Heron wrote:Trem apparenty has http download (since only very recently) . Can't say if they did that just for themselves, or whether its ioquake3 side (or even whether I got mixed up and misunderstood things :/...)

BTW @SeanHeron, ioquake3 has had HTTP Downloading for a long, long time.

xplinux557 wrote:Also, the physics are a bit less realistic that iq3.

I wonder though if this is because of engine limitations or game design preferences?

Flying Steel wrote:

xplinux557 wrote:Also, the physics are a bit less realistic that iq3.

I wonder though if this is because of engine limitations or game design preferences?

Do you mean player physics? Then that's game design preference, as you can easily change to VQ3 physics for example by using the following console command:

Code: Select all

exec physicsQ3.cfg


I have to say that I myself prefer VQ3/CPM/Nexrun physics over the Nexuiz ones, and hope there will be more servers that run these physics sets (or at least give the players an option to vcall them)

FruitieX wrote:

Flying Steel wrote:

xplinux557 wrote:Also, the physics are a bit less realistic that iq3.

I wonder though if this is because of engine limitations or game design preferences?

Do you mean player physics? Then that's game design preference, as you can easily change to VQ3 physics for example by using the following console command:

Code: Select all

exec physicsQ3.cfg


I have to say that I myself prefer VQ3/CPM/Nexrun physics over the Nexuiz ones, and hope there will be more servers that run these physics sets (or at least give the players an option to vcall them)

In ioquake3-based games, most of the objects' and players' movements are more realistic than in Nexuiz (generalization). In Nexuiz, whenever I shoot another player from the back, sometimes the player falls backward and not forward, and sometimes when I shoot from the front, the fragged player falls forward and not backward! This NEVER happens in ioquake3.

When I frag a person in midair in Nexuiz, the player's gun falls straight down and lands on the floor. In nearly all iq3-based games, when a player is fragged in midair, the player's gun falls at an angle and slides a little along the floor before coming to rest. Also, the motion of the player being hit by a gunshot and falling is more accurate in ioquake3 (my opinion). In OpenArena (using ioquake3), when a player is on a long stairway and jumps down several stairs, the player staggers a bit down a stair or two as it lands, just like a real human being regaining his/her balance (very pretty effect). The gibs in the iq3 gore effects appear to bounce a bit more than the Nexuiz gibs (iq3's blood is a bit disappointing, though). I think that some of the players' knees just buckle and fall over when killed in close contact (gauntlet or shotgun) and knocked over violently when somebody shoots them from far away (railgun or rocket launcher), but I'm not sure yet. This may or may not be true, seeing as I need to do some more observations when fragging in OA.

The only physics in DP that is more accurate than iq3 is the Darkplaces' shell impacts (shells ejected out the side of a gun and bouncing on the floor).

PS: I have confirmed everything that I've said in the last few posts here because I am constantly opening up Nexuiz and OpenArena to check my accuracy of what I wrote.

PPS: Does iq3 have ragdoll physics? (I'm just curious)

xplinux557 wrote:

FruitieX wrote:

Flying Steel wrote:

xplinux557 wrote:Also, the physics are a bit less realistic that iq3.

I wonder though if this is because of engine limitations or game design preferences?

Do you mean player physics? Then that's game design preference, as you can easily change to VQ3 physics for example by using the following console command:

Code: Select all

exec physicsQ3.cfg


I have to say that I myself prefer VQ3/CPM/Nexrun physics over the Nexuiz ones, and hope there will be more servers that run these physics sets (or at least give the players an option to vcall them)

In ioquake3-based games, most of the objects' and players' movements are more realistic than in Nexuiz (generalization). In Nexuiz, whenever I shoot another player from the back, sometimes the player falls backward and not forward, and sometimes when I shoot from the front, the fragged player falls forward and not backward! This NEVER happens in ioquake3.

When I frag a person in midair in Nexuiz, the player's gun falls straight down and lands on the floor. In nearly all iq3-based games, when a player is fragged in midair, the player's gun falls at an angle and slides a little along the floor before coming to rest. Also, the motion of the player being hit by a gunshot and falling is more accurate in ioquake3 (my opinion). In OpenArena (using ioquake3), when a player is on a long stairway and jumps down several stairs, the player staggers a bit down a stair or two as it lands, just like a real human being regaining his/her balance (very pretty effect). The gibs in the iq3 gore effects appear to bounce a bit more than the Nexuiz gibs (iq3's blood is a bit disappointing, though). I think that some of the players' knees just buckle and fall over when killed in close contact (gauntlet or shotgun) and knocked over violently when somebody shoots them from far away (railgun or rocket launcher), but I'm not sure yet. This may or may not be true, seeing as I need to do some more observations when fragging in OA.

The only physics in DP that is more accurate than iq3 is the Darkplaces' shell impacts (shells ejected out the side of a gun and bouncing on the floor).

PS: I have confirmed everything that I've said in the last few posts here because I am constantly opening up Nexuiz and OpenArena to check my accuracy of what I wrote.

PPS: Does iq3 have ragdoll physics? (I'm just curious)

I don't think it has. Also, we have a lack of animators, so there's no way to get a good animation where the player falls on his knees for example. And yeah, I'd like it if the dropped items/weapons fly away if the player was killed by some huge blast

what about darkplaces vs goldsrc? just a question, and i have goldsrc.

Interesting info on the graphics speed (though I would guess this might have to do with model level of details, and Level R-speeds or such as well).
I would think that most of the Physics stuff you're referring to is not engine necitated though - probably just a question of stuff being implemented in the game or not.

Could I ask you guys to make less heavy use of the Quote feature? perhaps quoting just the new stuff the guy before you wrote? Or even !shock! not quoting if you're simply referring to the post immediately above you ? It get's difficult to follow/read the way you've been doing it.

Mizu, on your question - how free/open source is goldsrc? Not at all I think, so I don't really see the interest in the question...

Now that id has been bought by Zenimax (Bethesda), I am crossing my fingers for a source release but wouldnt bet on it anymore. Plus it will be 2010 at the earliest, since it will have to wait until Rage ships (Carmacks own words).

I dont think Darkplaces is quite on par with id Tech 4 yet. Its certainly getting there, and it may have surpassed tech 4 by the time id open sources their engine. Id tech 4 is wholly geared towards realtime lighting, and although slower than a lightmap solution, it tends to look better. Doom 3 allows you to project shadow masks onto every surface simply by setting a .jpg key in the editor, whilst still having the light be wholly dynamic. This allows for things like a swinging ceiling light that projects a "fake" soft shadow cast by the grating onto the floor, as well as real stencil shadows if you stand under it.

Darkplaces realtime shadows seem to be very slow.

I have yet to see a Nexuiz map made wholly with blender. Another poster told me it was possible, but the lightmap resolution went down the drain if a single triangle was too big. Being able to model and texture an entire map in Blender has its advantages.

Doom 3 has ragdoll and prop physics. You can specify elasticity, weight, friction etc. for props. Not that thats really necessary in a deathmatch fps.

One thing Darkplaces does have is true hdr, whereas Doom 3 only has bloom.

SeanHeron wrote:Mizu, on your question - how free/open source is goldsrc? Not at all I think, so I don't really see the interest in the question...

Ask Valve/Steam as they no longer support goldsrc but still sell the games powered by it, ie its first game HL and up until HL2/CSS.

whitewolf wrote:Now that id has been bought by Zenimax (Bethesda), I am crossing my fingers for a source release but wouldnt bet on it anymore. Plus it will be 2010 at the earliest, since it will have to wait until Rage ships (Carmacks own words).

I was at his keynote and got the impression he would like to, so it comes down to approval - I can't participate in speculation as I am an id contractor.

whitewolf wrote:I dont think Darkplaces is quite on par with id Tech 4 yet. Its certainly getting there, and it may have surpassed tech 4 by the time id open sources their engine. Id tech 4 is wholly geared towards realtime lighting, and although slower than a lightmap solution, it tends to look better. Doom 3 allows you to project shadow masks onto every surface simply by setting a .jpg key in the editor, whilst still having the light be wholly dynamic. This allows for things like a swinging ceiling light that projects a "fake" soft shadow cast by the grating onto the floor, as well as real stencil shadows if you stand under it.

You can do such projection tricks with cubemap filters on lights in DarkPlaces too, it's actually more powerful.

I'm not sure it's fair to compare DarkPlaces and idtech4 on features, they are geared toward different things and have substantially different team size (one or two active programmers vs a dozen).

whitewolf wrote:Darkplaces realtime shadows seem to be very slow.

Actually the lighting is slower than the shadows, but both hurt severely, they have an extremely variable impact on framerate (the game runs very well all the time if you have realtime lighting turned off).

whitewolf wrote:I have yet to see a Nexuiz map made wholly with blender. Another poster told me it was possible, but the lightmap resolution went down the drain if a single triangle was too big. Being able to model and texture an entire map in Blender has its advantages.

Well subdividing a mesh isn't any big trouble in Blender, so I'm not sure what to make of this.

However all lightmap problems are in q3map2, I wish there was someone willing to upgrade/maintain that.

whitewolf wrote:Doom 3 has ragdoll and prop physics. You can specify elasticity, weight, friction etc. for props. Not that thats really necessary in a deathmatch fps.

Indeed, not necessary, partly because (unlike HalfLife2) the continual rocket-fire typical of a Nexuiz game would ensure the props are continually bouncing around the rooms and getting in the way.

As for Doom3, it's worth noting that the entire physics engine is in the released (moddable) gamecode SDK, it is not part of the engine, although the collision routine it relies on is.

whitewolf wrote:One thing Darkplaces does have is true hdr, whereas Doom 3 only has bloom.

I'm not sure it makes sense to compare hdr and bloom features as a deciding point on an engine choice

I do have much respect for what you've done with Darkplaces, even if it doesn't come accross that way

You can do such projection tricks with cubemap filters on lights in DarkPlaces too, it's actually more powerful.

This is very interesting. Do you mean for lightmapping? Could you maybe post a screenshot?

Also, I asked a question on this board before but didn't get an answer. Lordhavoc of all people would know the answer: is artist control over the hardness value of the specularity possible? Or is it a global variable hard coded into the engine?

I think the poster I was referring too meant that if one triangle was to big or stretched in relation to all the other then the lightmap would be messed up. I remember from modding jedi Academy that q3map2 doesn't play nicely when using models for architecture.

But yeah, keep up the good work LordHavoc

whitewolf wrote:

You can do such projection tricks with cubemap filters on lights in DarkPlaces too, it's actually more powerful.

This is very interesting. Do you mean for lightmapping? Could you maybe post a screenshot?

No it's for realtime lights.

http://icculus.org/twilight/darkplaces/ ... estmap.jpg

That texture is vertical stripes on the sides and top and bottom (it's a cube, like a skybox), there is only one light doing that, and there is no geometry to cast those fake shadows.

You can also apply any skybox to an rtlight as a cubemap to make a projector of scenery - it looks quite funny

The simplest example is that you can render a view inside a lantern model in a modeling app, and then apply that as a cubemap to the light and it does discolored lighting and shadows of the bars of the lantern and top and bottom, and such.

I don't have a better screenshot handy, but it is a way to texture lighting.

whitewolf wrote:Also, I asked a question on this board before but didn't get an answer. Lordhavoc of all people would know the answer: is artist control over the hardness value of the specularity possible? Or is it a global variable hard coded into the engine?

It's the cvar r_shadow_glossexponent in the console, however this applies to all textures, per-material could be added, per-pixel could be added (using alpha channel of gloss texture as an index into another texture for the gloss gradiant - but this would cause banding on ATI I fear).

whitewolf wrote:I think the poster I was referring too meant that if one triangle was to big or stretched in relation to all the other then the lightmap would be messed up. I remember from modding jedi Academy that q3map2 doesn't play nicely when using models for architecture.

Someone needs to make it play nice.