People respond to emotional content in communication not specifically in kind, but at level of intensity. If a conversation starts out unemotional, others will respond in kind. If someone breaks the wall and raises emotional language, other people will allow emotional language of that intensity to enter the conversation. The nature of the emotion (pleasure, anger, friendliness) is not always echoed.
The combination of these two suggests why flames and other forms of conflict are rife on the internet: a positive feedback cycle gets started. For instance, someone posts something that has a little humor in it. This is misinterpreted as having some irony in it, and this allows the reader to respond with irony. This in turn is thought to be be sarcasm, not irony, and allows readers to respond at that level of emotion. Sarcasm is then misinterpreted as mild-insult, and then to insult, and then to anger. Now the emotional intensity is quite high; thus the flame.
Therefore, dampen the emotions you use when posting online. First, AssumeGoodFaith and assume as little (negative) emotion as you can on the part of a poster. Presume that sarcasm is probably irony, and the rudeness just brevity. (TrustButVerify.)
Second, ModelDesiredBehaviour and keep text impersonal. It may help to know that certain words have more emotional content then you may intend. "Assume", "didn't", "should" and "forgot" have a lot of emotional baggage. Try to start with a "yes" rather then a "no" - "no" can have more emotional content when presented first. Avoid "blame" or "responsibility" phrases. Refer specifically to your own emotional state, with smilies, or with an explicit statement ("I'm not angry, but just a little concerned").
You are responsible for your own internal emotional state, not that of others. Your emotions are not a curse inflicted on you by a dark God.