hm... ever since i started playing with gl_texturemode 3, i can't stand the sight of the blurred bilinear filtered textures. :P 

I'm afraid my e1m2 retexture attempt ( http://www.quaketastic.com/upload/files/screen_shots/e1m2_tex3.jpg ) died a death when my laptop got stolen.

Glad you enjoyed the idbase textures though, I was very happy that the response from everyone here was so positive, considering most people (like me) are predisposed against retexture packs. 

that looks AWESOME. wow. fully high res yet unmistakably the original texture!

real shame about the theft. :( 

would have loved to see that. 

that you have to make daily backups! Sorry to hear about the laptop, that picture looks really great. 

and yeah... lesson learned :/ 

FitzQuake 0.85 @ 1024x768 - 16xQ CSAA.
~
gl_flashblend 0
gl_polyblend 1
gl_overbright 1
gl_overbright_models 0
gl_texture_anisotropy 16
gl_texturemode 3
host_maxfps 120
r_lerpmodels 1
r_lerpmove 1
r_particles 2
r_shadows 0.4
r_wateralpha 0.4
scr_clock 3
scr_showfps 1
scr_sbaralpha 0.5
vid_refreshrate 85
vid_vsync 1 

hm... ever since i started playing with gl_texturemode 3, i can't stand the sight of the blurred bilinear filtered textures. :P

Ha ha, I know the feeling. Compare health boxes with and without filtering - without is just so much more crisp and clean.

Welcome to the club. :) 

that just linear_mipmap_linear? 

GL_NEAREST

or maybe something like GL_NEAREST_MIPMAP_NEAREST ?

uh, well, unless linear is the same as nearest? i don't know much about the terminology 

http://www.quaddicted.com/software-vs-glquake/software-vs-glquake-texture-filtering/ 

GL_NEAREST_MIPMAP_NEAREST should be preferable as software Quake used 4 miplevels. For best accuracy the engine should also restrict mipmaps to at most 4 levels too. With GL_NEAREST you're going to be getting a lot of noise and aliasing artefacts on far off surfaces. 

if I understand this right, GL_NEAREST_MIMPMAP_NEAREST increases pixelisation but also detail? 

Generally GL_LINEAR and GL_LINEAR_* filters will blur. GL_NEAREST and GL_NEAREST_* will give you pixels.

GL_NEAREST will give you a lot of flickering and ugly patterns.

The differences between GL_NEAREST_MIPMAP_NEAREST and GL_NEAREST_MIPMAP_LINEAR are almost invisible. Try the outside on marcher.

I guess the first part of the string is the non-mipmapped texture and the ending is for the filtering of the mipmaps? 

just make sure to turn on anisotropy for mipmaps! it makes a world of difference and unlike linear and bilinear filtering, works very well with old school low-res art. 

In fact I have no idea if I have that enabled. 

r_anisotropy 16 to set it to highest in fq/qs.
some engines don't have it (or the command is different) but you can always force this kind of setting in your ati/nvidia driver config menus. 

gl_texture_anisotropy in fitzquake (and darkplaces i think) 

Since all my maps use QRP textures, and I also have them installed for id1, I always use GL_LINEAR_MIPMAP_LINEAR. It just looks best to me. 

gl_texture_anisotropy in fitzquake (and darkplaces i think)

oops, i totally got that one wrong. :x 

have been using GL_LINEAR_MIPMAP_LINEAR with 16x anisotropy since forever. 

I guess the first part of the string is the non-mipmapped texture and the ending is for the filtering of the mipmaps?

Not really, Quake's options are a bit misleading as they don't really reflect the way texturing works.

You've got two basic types of filtering on a texture - what happens when the texture is magnified and what happens when it's minified (a third relates to how miplevels are blended). The setting as exposed in Quake is just that for minification, but a magnification filter is kinda derived from it.

So, GL_*_MIPMAP_* - what does it mean? The second part refers to how an appropriate mipmap is chosen, and the first refers to how the correct texel is chosen from that mipmap. GL_LINEAR_MIPMAP_NEAREST for example will just pick the closest miplevel in size for the pixel being textured, then take a weighted average of 4 texels from that miplevel. GL_NEAREST_MIPMAP_LINEAR picks the two closest miplevels, then takes a nearest neighbour sample from each, and produces a weighted average of those two samples for the final result.

It's quite legal to have a magnification filter of GL_LINEAR but a minification filter of GL_NEAREST (or GL_NEAREST_MIPMAP_*) - Quake just doesn't allow it. Instead it grabs the first part of the minification filter and forces you to use that. 

Wow, after carefully reading this twice today (I think) I actually understood it. Thank you very much!

While we are on the topic, how does one change the texturemode in DirectQ (if that is possible)? I saw some gl_ variables (and was surprised and thankful for I guess the config compatibility) but none for texturemode. 

gl_texturemode is there, I've kept the same names as OpenGL uses for it (for consistency), it has TAB autocompletion on ("gl_texturemode gl"+TAB - >grin<) and is also accessible in the video options menu (although it's called "texture filter" and "mipmap filter" in there). 

Oh, silly me was using version 1.3 without noticing... 1.8.8. sadly does not run in Wine (failed to create smokepuff texture) and my laptop has no Pixel Shaders so no go for me. BUT I just needed to know if and how, so thanks!