Blending and clamping
According to glBlendFunc all scale factors have the range [0,1] and according to glBlendEquation all values are clamped to [0,1].
I can understand this if the image format is GL_RGBA8, but not if it is a floating point format. Have I understood this correctly?
If so, what is the reason?
It makes it harder if you want to do HDR imaging.
Don't blame me; that's how the man pages have always described it.
Those descriptions are inaccurate. The clamping behavior only exists when writing to non-float-point or non-integral targets. And division by whatever also only matters when dealing with fixed-function.
In short, they do what you think they do, not what they say. Feel free to update the pages; that's why they're on the Wiki.
Thanks for the confirmation, I updated the Wiki.