The biggest problem in Computer Science was the fact that people care about the technology. The physicists did not care whether the system they were using supported the concept of transclusions, or whether it used a self-describing metadata format, or even if it could be proven to scale across a distributed system. They were pragmatic about it, because they're physicists not computer scientists and so the computers were just a tool.
Whereas for a Computer Scientist the open access technology itself is a potential research topic. So you get crazy stuff like a project to figure out how to perform searches across potentially hundreds of OA repositories in a distributed system, all of them with separate policies and metadata formats - instead of one working repository.
On the other hand, once things started to take off, this turned into an advantage. Can't get administrative funds for the much needed performance optimisations in your archive containing 25 years of data structures papers? Make it a research project and get a grant. This seems to have worked out OK for e.g. Southampton.