Render motion blur vs add in post

[art3mis]

It's possible to add camera effects like motion blur, depth of field in post. Doing so gives you much more flexibility.
But of course its easy to add these at render time.
My question is, when does it make sense do render motion blur, dof, vs add them in post?
Are there any guidelines or best practices?

[haggi]

You can do dof and motionblur in post but it is important to keep some points in mind:

• Motionblur will always be linear, no problem in most cases until you have fast curved animations.
• both effects will not appear in secondary rays (reflections, refractions, transparency) what can lead to strange artefacts.
• doing both in post can be tricky because with motionblur you blur the z channel or vice versa what can create another artefact.
• Dof reveals geometry in the blurred areas so if you try to create dof with only one image and a z channel you will have problems because you try reveal geometry which does not exist.
• Same with motionblur, the geometry will be blurred what means:

In both cases it works better if you can separate the layers, e.g. the moving object and the background. I'd recommend to do only one effect in post and I'd do dof in post.