The At&t decision seems flawed to me. The disctinction the court makes between abstract software information and a tangible copy of software seems irrelevant. Microsoft didn't export "abstract information" it exported a tangible copy, i.e. the golden master disk. The court's distinction seems to be that if Microsoft calls the disk a "master disk" it's information, but if Microsoft calls the disk a copy of the Windows operating system then it's a tangible copy. Or maybe the court is saying that the information that was actually transfered from the master to the computer was abstract and therefore not a component, but if that's true then Microsoft would not have been guilty of violating 271(f) even if they exported each copy of Windows individually, because the intangible information on the CD would still have to be copied from the CD to the computer. In other words, the CD would not become a part of the computer, rather the intangible information on the CD would be transferred to the tangible form on the computer. I find it hard to believe that the issue of whether software is a component of a patented invention depends on whether one makes several copies of the software from one disk, or whether several disks are used to make one copy each.
Software is a blueprint no matter how you look at it. The confusing part is that we've developed machines that can actually "read" blueprints.
For an interesting case note on At&t v. Microsoft see 59 ME. L. REV. 191 (2007). (I should know I wrote it.)