Coase argues that firms grow as long as the costs of organizing transactions outside the firm, using the market pricing mechanism, are smaller than the costs of organizing those same transactions inside the firm, using management.
Today we are seeing a trend toward running more and more of our networked applications inside of Google's data center. (Or Amazon's, or whatever.) Perhaps this happens because the costs of organizing exchanges of information between administrative domains on the internet --- not necessarily using a market pricing mechanism, but not involving complete mutual trust --- are nonzero, and the costs of organizing those same exchanges inside of a single administrative domain are sometimes smaller. Suppose this is the case. What can we predict? What kinds of technological or social developments might increase the costs (or decrease the efficacy) of exchanges of information between administrative domains? Spam is an obvious example; so is software insecurity in general. Proprietary systems software, or systems software that is not designed for internet scale, could also have such effects. What kinds of technological or social developments might decrease the costs or increase the efficacy of these exchanges of information? Better systems security, better decentralized source control systems (so that software improvements can be shared more readily between administrative domains, standardized protocols covering a wider range of the needed functionality. But the explanation for these trends might not be that inter-domain costs are going up; it might also be that intra-domain costs are going down. Things like the Google Filesystem, Mapreduce, rsync, backuppc, Xen, VMWare, and the like, may allow larger systems to be run by fewer sysadmins. (Riffs on a discussion tonight with Rohit Khare, who's visiting us here in Argentina.)
