You are trying to help the authors-not to show how smart you are. Be constructive. The more constructive you are, the more the author will get from the discussion, and the more likely it is that he/she will help you in the future. It’s a small world. The more helpful you are, the more likely that people will figure out that you are smart, anyway.
Do not say ‘This is a great paper,’ or ‘I really like this paper’ and so on. Everyone says it, and therefore it means nothing. Don’t be phony. Be helpful.
If you find a problem in the paper, or you think the paper is really stupid (it can and will happen), clearly and calmly explain the issues. Do not ignore them. But whatever you do, try not to make the author look stupid; see point 1. Telling the audience calmly but confidently the problems in the paper shows that you are a serious and thoughtful person. That’s what you want.
Be prepared with professional looking slides.
Don’t go over your time. Be short.
If you can add some intuition to the paper, do it. The author always runs out of time.
If you can bring additional information, data, or examples to the paper, do so. I have seem some great discussions in which the discussant redid some of the analysis with new data, or used a slightly simper analytical model to expland and explain the paper’s results. Those discussions surely saved the author a round or two at a journal.
Be creative in expanding the scope of the work.
Don’t be an asshole, only the big shots can do that successfully. And you never know which person in the room will be reviewing you for something in the future.