I am writing a game in XNA 4 and I want to have something like a corridor. Do you think that is better to use one big and solid model, or to use multiple small parts (window, stairs, door...)?
In terms of graphics rendering, one large mesh is faster to draw than several smaller meshes.
If you're talking about something like a corridor, it's probably not going to be that intensive for the GPU if you send it in as a door and a floor and some walls. The smaller meshes also allow you to fiddle with positioning more easily.
If your drawing becomes a bottleneck, then consider it.
I'd suggest keeping the door as a separate model, unless you're certain you never want to be able to open it.
Windows if they are transparent also want to be separate models so they can be depth sorted appropriately for rendering.
Everything else should probably be a single model and texture for best performance, and to eliminate any cracks where they join together.
in most cases using small parts are worth it because it gives your code reusability and flexibility; and in case of a corridor or something similar that does not include MANY (many would mean a number like at least 40) smaller parts, the performance gain of using a big and solid models is negligible(except if you are drawing the object many times per frame).
if you care about the performance i recommend precomputing the lighting or using some kind of vertex arrays(vertex buffers?) which would actually make a difference. ( if those things are possible in XNA - i dont know a lot about XNA )
p.s. : the first paragraph of this post is correct only about drawing ( not about for example using separate physics models for parts of the corridor )