It all comes down to one critical point; If you don't think you can sustain the pace necessary to meet the schedule Sales promised, then do not commit to the work. If Sales overpromises, it's not your problem, until you agree to the work. THEN it's your problem. Remind your boss that all the sales guys have to do is say yes and they get the check; you're the one that actually delivers on the promises, so if you say it won't work, your boss should be listening. If you have the ill fortune of having a manager who listens to his sales force more than his development force regarding what is and isn't possible, then you have the Dilbert-esque PHB, and you should be updating your resume.

This is one reason I like Agile; the development team is involved in the process from the initial design discussions. You can calibrate a "point" from both ends; the dev team decides (either explicitly or empirically) roughly how much development man-hours are inherent in a point, which management can then use to calculate points per week, points per month, etc which leads to a dollar figure. At that point, your sales team now has figures relating to the cost and time required for the current staff levels to get the current amount of scope done. If they overpromise once they have those numbers, they are out on their ass.