cl_himodels should be deprecated/removed

[ghost]

cl_himodels is a cvar that allows the use of "high quality models" back from the WON era. Basically player models had an alternative submodel which included more polygons.

This feature should be dropped because steampipe's HD Models option serves the same purpose (albeit, currently used in Counter-Strike) and doesn't limit playermodel submodel customization in custom server-side mods.

[ChaosSpoofer]

I don't know about CS, but cl_himodels still applies to HL/HLDM player models, which aren't impacted by the Blue Shift/SteamPipe HD model option (and not everyone's a fan of the HD replacement set regardless).

[ghost]

IIRC HLDM and OP4 use this cvar, although no playermodel in OP4 includes an extra polygon submodel, it's probably there for legacy. Team Fortress uses a similar cvar only used for that game that actually changes the player model instead of changing the submodel, which made sense in the WON era. (TFC should probably be included in this issue as well).

Going back to HLDM, the best case scenario would be removing the lower quality submodel (reasoning for this below), and add the PS2 Player Models in valve_hd although the latter is unlikely.

There's literally no reason to keep the lower polygon submodels anymore, given on how Steam is required to play the game and the program actually exceeds the game minimum requirements, specially with the upcoming steam client beta which uses Electron.

As stated in the OP, removing this cvar would allow more server custmization, letting operators customize the player's submodel value, which is currently handled client side.

[ChaosSpoofer]

As a hardcore TFC player, I can tell you that the old models are absolutely preferred by the competitive players, as they're more accurate to the hitboxes and the animations are easier to read. Beyond that, it's a cosmetic choice between which one players prefer. There's no reason to eliminate that choice from players.

As I said before, in terms of HD HL models, a lot of people still prefer the charm of the originals as well, since the HD ones radically alter the designs of some of the enemies/weapons, etc.

To an extent, I can see cl_himodels exhibiting similar issues.

Here are some of the differences between the two Gordon models, as an example:

https://i.imgtc.ws/EgNMxI2.png (0) https://i.imgtc.ws/MrwALwt.png (1) https://i.imgtc.ws/SA7pHMQ.png (0) https://i.imgtc.ws/Cr8a1yM.png (1)

The original models are clearly bulkier than the higher-poly ones. I imagine one set corresponds to the hitboxes better than the other, and as the lower-poly ones were created first (or are otherwise the default), my guess would be they're more accurate. Similar issues exist with TFC's two choices, where as I mentioned before the older, "uglier" models are far more accurate to the hitboxes, while on the newer ones you might think you're aiming at the top of a player's head, but you're really aiming at the hitbox's neck region and therefor won't score headshots. It's a real issue. So the lower-poly models might be more faithful to hitboxes and offer competitive advantage.

Secondly, there are cosmetic differences with the character. You see the higher-poly one adding a rat tail to Gordon, which is pretty ugly IMO, and he has a weaker-looking face. I haven't tested the others atm, but I seem to recall noticeable changes between the other available models, as well. I can see the differences being a cosmetic choice for those who really choose to care about such things, or prefer the nostalgic charm of the originals.

And finally, GoldSrc models still do come at a cost even in the modern age. I have an i7 6700k and GTX 1080, and in TFC, I can maintain 2,000 FPS staring at a wall, but I can barely manage to maintain 240 FPS on a full server with everyone onscreen at once. And I want to maintain 240 for my 240 hz monitor, so for TFC I do specifically choose less-demanding models over higher-demanding ones.

For all of these reasons, I see no reason to strip the choices and the commands from players.

If server customization is an issue, I'd address it from the standpoint of the server itself. If the server can't set the cl_himodel value (due to cl_filterstuffcmd or whatever), then my own personal choice of solution would simply be to include a line of text (in motd or piped onscreen) instructing the player how to change the value manually so as to take full advantage of server-side customizations.

[ghost]

All those are fair points of concern based on the request, however, I think you might misunderstand what did I mean.

I want to emphasize that I do not wish restrict player customization or let server operators change the player's choice.

Beyond that, it's a cosmetic choice between which one players prefer. There's no reason to eliminate that choice from players.

And I want to maintain 240 for my 240 hz monitor, so for TFC I do specifically choose less-demanding models over higher-demanding ones.

I said before that I don't wish to remove that choice from players, under the new proposal, the "new" TFC models would go into tfc_hd (currently TFC doesn't have a Steampipe HD model set) while the default option would be the original TFC models. Which a lot of old timers wouldn't notice the change because they have ticked off the "Turn on HD models" options to bypass the grenade.mdl consistency bug.

This is also way more straight-forward for the user, instead of going to advanced options.

The original models are clearly bulkier than the higher-poly ones. I imagine one set corresponds to the hitboxes better than the other, and as the lower-poly ones were created first (or are otherwise the default), my guess would be they're more accurate. Similar issues exist with TFC's two choices, where as I mentioned before the older, "uglier" models are far more accurate to the hitboxes, while on the newer ones you might think you're aiming at the top of a player's head, but you're really aiming at the hitbox's neck region and therefor won't score headshots. It's a real issue. So the lower-poly models might be more faithful to hitboxes and offer competitive advantage.

Secondly, there are cosmetic differences with the character. You see the higher-poly one adding a rat tail to Gordon, which is pretty ugly IMO, and he has a weaker-looking face. I haven't tested the others atm, but I seem to recall noticeable changes between the other available models, as well. I can see the differences being a cosmetic choice for those who really choose to care about such things, or prefer the nostalgic charm of the originals.

The lower quality player models were made to reduce FPS drops in hardware of the era, in situation of 32 player death matches.

The higher polygon submodel represent closely to the actual in-game models. The best example of this is the gordon player model you mentioned, just compare the higher polygon gordon model with player.mdl, they are almost the same (including ponytail, which is, in fact, also added to the PS2 version of Gordon), but the player model has less polygons and it has color support.

It's not something of being "closest" to the hitboxes, that's just a placebo effect, otherwise everyone would be using cl_himodels on 1 because that causes missing player models to display a helmet on gordon (player.mdl), causing it to cover more area of the head hitbox. Ironically enough, the helmeted model has actually less polygons.

gman, zombie, robo, recon and hgrunt do not have a higher polygon variant.

If server customization is an issue, I'd address it from the standpoint of the server itself. If the server can't set the cl_himodel value (due to cl_filterstuffcmd or whatever), then my own personal choice of solution would simply be to include a line of text (in motd or piped onscreen) instructing the player how to change the value manually so as to take full advantage of server-side customizations.

As said above, it's not matter of servers restricting player choice, but to standarize GoldSrc games to use actual SteamPipe features, and also, to get around current engine limitations. If I want to represent player states with a custom model that would mean I have to use a valuable precache spots by precaching a custom model I need, imagine I want to create a mod that changes the playermodel on each state, I do not have to precache 1 but n times I want to do it, instead of just precaching 1 model with all the stuff I need in submodels. Which is huge considering the small limits we have.