CI systems only need the "centralized" copy as a reference point of what the current master is. Everyone gets their own entire copy of that repository though. If the truth suddenly changes it's known. I also know that if the copy in the "centralized" matches my own then the CI is also running exactly my code. If the centralized machine fails I can fail it over to any copy on any machine. That seems pretty distributed to me as compared to the original "distributed" version: e-mailing source code around. Version control like SVN and CVS were centralized in that me having a copy of the source and you having a copy of the source indicated nothing about whose version is correct. Moreover a failure of the central spot in the system is fatal unless you had backups as a form of pseudo-distribution whereas in today's version control a failure on any one machine is irrelevant to the preservation of the system.