GL_NV_depth_buffer_float provides a format of texture that can store depth buffer information not only in 32-bit but also in float. This is for use with a FBO (framebuffer object)
In your case, you want to deal with the default/system framebuffer.
You have to request this format from the system.
If you use glut, it is not exposed, so you cant do it.
if you are under Unix using glx, in the FBConfig you can request
GLX_DEPTH_SIZE to be 32. (check with glxinfo the formats that are actually supported)
If you are on windows, using wgl, deal with the pixel format descriptor, and set cDepthBits to 32.
If you are on mac, using Cocoa, set the value of NSOpenGLPFADepthSize to be 32 in the NSOpenGLPixelFormatAttribute structure: