I have one architecturally significant comment to this document. The use of iCalendar to schedule device services (for example the specification identifies Demand Response as a possible application) across a large domain of buildings (or equipment within a building) could have the unintended effect of load synchronization. For example, if a DR event is delivered with iCalendar to 1,000,000 buildings and set to end at 5pm, there is the possibility that the entire reduced load comes back online at exactly 5pm, causing a demand spike and potential grid instability. I would have thought that there would be a randomization interval provided as an optional element of the Interval time segment (or part of the performance attributes) to prevent this type of occurrence.