In Quake you can enter a room, shit yourself, and back out of the room to re-evaluate your strategy in the knowledge that a big conga line of monsters is not going to just perfectly chase you down all through the map.

Any enhanced pathfinding AI needs to appreciate that this sort of soft "tethering" of monsters to an area may be just a quirk of crap pathfinding at the moment, but monster area-limiting would certainly need to be a feature available to the designer in any future AI rewrite. But it can't be rigid, because you'd still want to be able to "pull" a monster to anywhere you want as well. 

He's spot on. AI needs a refresh. I'm sure there's some way it could be offloaded to the compiling stage like Q3A. So the mapper adds some entities to help with dumb monster behavior and then an external solution takes that data and bakes it into the map. Engines devs would then have the option of supporting the new AI feature or not.

Would this be feasible?

I appreciate the convenience of a mouse driven menu but frankly I have a high DPI mouse that always runs too fast in these types of UI. It's certainly welcome and I am all for making things more usable for noobs.

I think it's important to remember that communities have to grow to succeed in the long run. We want Quake as appealing and accessible as ever. I still think it's possible to create a set of standard that can be voted/agreed upon and then collaborated on by the brightest minds in our coding community. 

So the mapper adds some entities to help with dumb monster behavior and then an external solution takes that data and bakes it into the map.

Q3 looks at the connectivity of the BSP leaves to generate the AAS graph right? Something like that sounds like a possible approach, rather than the designer having to add loads of nodes everywhere.

Of course, once you have perfect pathfinding, it's then up to the designer/modder to write a load of QC code to do all the extra AI logic that stops the whole level playing out like a Benny Hill sketch. 

Honestly, advanced AI belongs to "what do you want to see in a mod", not "what do you want to see in Quake". AD is the perfect candidate to such a thing, and it already has tethered waypoints AFAIK. 

I think my breadcrumb solution is reasonably "ok"... it would mean if you were properly keen on running away you should still be able to do so.

the idea would be:

every 5 seconds or so the player spawns a short lived invisible entity (breadcrumb), if the monster can't see the player, it searches for that breadcrumb, and travels to its location if the breadcrumb has line of sight to the player.

If the breadcrumb doesn't have line of sight, the monster just continues on its usual dumb pathfinding.

this way the monster is likely to get an extra corner or so of pathfinding before losing the player 

Something along those lines could work. Not an invisible entity, but just a .vector field (self.enemyorigin?) containing the location where the player was last seen by the monster.

Since the sight line is a straight line, and monsters only navigate properly across straight lines, this should fit their logic.

Actually, this should be dumb easy. I'll take a look at the QC code now. 

Hmm, that would require a dummy entity anyway, as movetogoal is implemented engine-side. 

calls em "player trails" 

I'd love to see certain enemies use others as meatshields... chunkier enemies up front and center while the smaller ones try and flank the player

I'd also love to see ogres figure out how to lob grenades over other enemies heads at the player 

rather than the designer having to add loads of nodes everywhere.

I'd rather have control over it though - as no matter how good the implementation is there will be issues. 

what I'd like to see would be engines to FINALLY start supporting tracebox, so that this stuff can actually be implemented...
without that, good luck with this ai stuff... you'll need it...

but yeah, without this being some map-specific thing, you're going to drastically change the balance of various maps. 

Breadcrumbs is the only AI "enhancement" that should be considered.

The direct-path, omniscient AI is actually quite good and varied due to the way Quake handles AI states.

AI in Quake is handled on a per frame basis, allowing perfect coupling of the animation with the movement and movements to match the model. This prevents monsters feeling too floaty when walking. The simplistic AI lets the player be smarter, gives smart players a way to control monsters and make in-fights, and gives some monsters character. I always love it wgen shamblers wander off for a bit becahse they don't see me.

Getting caught on lips is bad though. 

Reading the DLL rant made me lose a few IQ points.

Something I'd like to see is an engine which focuses on a modern tool-chain with modern formats. No one is considering Quake engines for bigger TC projects because every project also seems to aim for compatibility and misrepresents itself and its capabilities.

When you look for ways to get started for mapping/modelling for modern Quake engines you still mostly get results for .MDL and Quake 1 BSP creation. I highly doubt most people (unless they are trying to cash in on the whole 90s FPS revival...) would want to work with their limitations. Most noobs stop when they realize that you can't crouch due to hull size limits. I GUESS EVERYONE JUST WANTS TO MAKE SIMPLE QUAKE MODS.

Engines just seem to do a little too much overall with their builtin HUDs, menus and scoreboards. If they don't support CSQC then mods have a harder time customizing the look and feel. Overall all the advanced engines are bloated hogs with an identity crisis.

Thoughts on currently relevant engines:
QuakeSpasm
To me was always a solid GL engine that just did NetQuake stuff well. I'd never consider it for development. Just for playing old NetQuake mods. Extending it to do more complex things might harm it in the long run, because it'll just be another FTE/Darkplaces.

Darkplaces
While it somewhat supports CSQC and somewhat supports Quake 3's material/shader system, doesn't do any of it well. It's only half QuakeWorld compatible as well. So as a developer I've run into a few problems.

FTE
Does everything and _most_ of it works. Sadly the engine just does way too much. 16 year old code sticks out and things regress more often, because it's all cobbled together in one giant binary, I often have run into bugs/features that hadn't been tested recently or at all. It's like the emacs of Quake engines. But probably more reliable than any GNU project...

ezQuake
Does anyone even consider this for development? I guess if you run a custom QW server you use it to test because there's this invisible community of QW players who seem to use it. I don't know anyone personally who does, but a bunch that have focus on retaining compat with those clients because half their player-base uses it.

Overall, all of the above aim to be Quake 1 compatible. Those who care about Quake and don't need obnoxious high-res texture-packs just usually stick with QuakeSpasm. Why? Because it focuses on one thing well. Technically, FTE is the most complex engine but it has less of an adoption than say Darkplaces because it's mostly competing with ezQuake in the QuakeWorld space. That's because everyone has a hard-time at comprehending what it actually is. The name suggests it's just another ezQuake alternative. That's where it's kinda stuck.

I don't follow QuakeSpasms development (I don't even use it), but I hope that those who maintain it know what they are doing. Don't give in to requests of adding support for non Quake file formats and conventions. Just stick to being a good NetQuake client that offers the feel of the original. Otherwise you're just a few inches away of supporting rtlights and giving little kids the keys to using their Rygel pk3s. It'll be like Darkplaces but with less features.

Having been busy in recent years, I just occasionally take note of what's happening in the engine world. So what I might be talking about is not relevant anymore. 

Does support colored lights.

I can agree on FTE. Not really sure where it fits in, no offense to Spike or anything. 

@Qmaster

FTE is a perfect engine for people like me who wants to play Quake with the classic look (pixelated textures and particles) but also can't live without realtime light and shadow.

Now, answering the question "What Do You Want To See In Quake In 2018?"

I just want to see mappers and players noticing how important realtime lights and shadows are for quake.

Seriously, it makes the ambience so much more gloomy and terrifying. Just look how this little and simple old map looks awesome with realtime L&S: https://www.youtube.com/watch?v=WmcXLNzan48

The dark atmosphere, the monster's shadows moving on the walls... reminds me of Severance:Blade of Darkness. That game blew my mind, the gameplay was slow, hard and boring, but I played it just for the lights and shadows, it was awesome runing around the map with a torch on my hand, moving shadows... it makes the place look real. And I wished so bad that quake had it too... And now we have, and looks like nobody care for it. It's awesome when you see a shadow walking at the wall before you see the monster, the shadow grows and shrinks, so you know when he is close or far away, or when you see a rotating shadow from a weapon that you can't reach... these little details add so much to the game. Imagine how mappers could create suspense, paranoia and horror if they think about RT shadows when they are making a map...


Sorry for the long post :P 

 

Well...I would have a leg to stand on by arguing that Quake needs point lights sprinkled throughout a space as fill lights, but...

ericw already made this method of lighting obsolete with his bounce lighting.

With too many point lights comes dozens of overlapoing shadows = ugly and distracting and poor performance...on older maps that are much more evenly lit. 

Something I'd like to see is an engine which focuses on a modern tool-chain with modern formats. No one is considering Quake engines for bigger TC projects because every project also seems to aim for compatibility and misrepresents itself and its capabilities.

So, you want an engine that works with modern asset formats, but you're also knocking points off engines for being compatible with the original Quake.

Ummm... maybe just do whatever in UE4 or Unity? Not sure I understand why anyone would even want a Quake engine in this situation... 

I've noticed just how out-of-place and incongruous real time lighting looks with Quake's old textures and models, especially with pixel mode on, hope that helps. 

I think it fits perfectly with them =D

but it's ok, let's agree to disagree ;)

i'm not here trying to change anyone's minds, i'm just answering the question of what i would want to see in quake in 2018. But nobody has to agree with me, it's just my two cents ;) 

These engines run on certain ranges of modern hardware, Unity struggles heavily on integrated chips and even dedicated cards from 5+ years ago, also have all sorts of licensing issues and do not contain the same workflow that everyone is used to in the Quake community.

There is a lot of tech in between QuakeOne-level-tech and something like UE4. When it comes to 3D Gaming engines it's probably as big of a gap as there can be. You never have to take full advantage of such things. I am sure that anyone would welcome a Quake engine where the audio, physics and rendering API was on par with something like the Source engine (because no one has $25000 to pay to Havok and RAD Game Tools)

But seeing as only one current Quake 1 engine fork (being FTE) even attempts to take advtantage of things such as a Bullet physics implementation or supporting OpenAL Reverb setups, that's not really happening. So that's what I'm talking about.

Of course, what I named above is an example. People love Quake modding, they are familar with it. Everyone here takes advantage of the GPL all the time and are able to build fantastically open things - but all everyone cares about is making it so that Quake 1's playable in 1000 new ways instead of opening it up to new concepts. 

aal everyone cares about is making it so that Quake 1's playable in 1000 new ways instead of opening it up to new concepts.

Yes, and? 

For what it's worth, engines probably typically want to react to what the community is actually doing, rather than create a brand new content creation paradigm and then hope that some modders pick it up and start doing something worthwhile with it.

DP, FTE have md3 and Q3 bsp, amongst all manner of other things, and have done for a long time. When we start seeing good projects based around those features, we might start seeing support in other engines. 

this thread title is like a legal contract - realistically. Its such a relative term :D I'll try.

Realistically I would love quake's palette restrictions lifted. I understand that its all comes down to the fact that old wads and bsps can't store truecolor images, but (correct me if I'm wrong) you're converting them to truecolor anyway when loading into memory?

So bsp can store palette.lmp. So what if, upon loading said bsp, engine detects it has palette.lmp and uses it for textures from that bsp, instead of default one? Obviously items and weapons would still use original palette.

From my ignorant perspective it sounds very realistic. But then again I have no idea about inner workings of Quake engine, let alone source ports.

Point is - using colors from original palette is very limiting. It might be a good thing to keep things coherent and 'in-style', but from what I've seen so far, people would rather create awfully translated to quake's palette textures than adjust their design decisions. The idea of creating a new mod just so I can change color palette is absurd and discouraging. 

Whoa, you can totally remove all palette restrictions right now by just using external textures! Works in every custom quake engine I know.

You can even keep your new fancy coloured textures looking all limited-palette and pixel arty, but that's an artistic decision now.