Which is better?

Me and my friends are on an argument regarding modeling for source. This question should be aimed to expert source modelers who know what to do and how to do it.


If you can, please also post the reason of your choice

1)Different parts intersect without sewing, and their polygons end one inside another, even on very complex shapes like weapons.

http://dl.dropbox.com/u/2951174/tube/1.jpg

2)All the intersecting parts (except those which only go one trough another for a little part and then get out) should be sewed together to avoid intersection

http://dl.dropbox.com/u/2951174/tube/2.jpg

Which of these techniques is better?

I’ve noticed many tutorials about modeling for source and basically all of the source models themselves use the second technique, but my friend says it does no harm to use the first, and it’s faster.

If I remember correctly, Source doesn’t like faces inside of other faces

why exactly?

Non-manifold. It sees it as a never ending shell and cannot resolve it.

And what happens because of that ?

Personally I prefer to use first technique. It will not make your model bad but you will have little problem with texturing later. Also, may I know what you’re modeling?

Prove it because what you’ve just said it’s heresy.

Weapons

Then 2nd technique will be good and better too.

Depending on the model. First method for miscellaneous world props with little detail, second for things that are detailed and that you can spare some polys for.

[editline]12:57AM[/editline]

Though you should only use the second method when you can actually make the pieces match edge-wise. If you can’t, use first.

[editline]01:00AM[/editline]

The plus side of the second method is that it results in better Gouraud shading. The negative side is that it creates unneccessary polygons.

[editline]01:13AM[/editline]

Wrong. The worst thing that can happen is some Z-fighting on mostly parallel surfaces.

Yup, totally depends on the model, they both have their purposes.

Really depends if it’s going to increase the triangle count and make it harder to UV map.
I use a combination of both.

I use first because it’s easier to UVmap, it saves polygons, it won’t look any worse and it’s nicer to look at on a wireframe.