A year ago I wrote that a press release with five logos does not lay fiber, and that the specs would tell me whether the O-RAN Alliance actually meant it. The first real specification landed at Barcelona this week, so I owe the story an update.

The deliverable is the Open Fronthaul specification, and it matters because it nails down the single hardest seam in the whole architecture: the link between the baseband unit and the radio at the top of the mast. This is the exact spot I flagged when xRAN first promised open interfaces, the place where latency goes to die and where earlier open-telecom pledges quietly turned into one vendor’s private dialect. The spec defines the control, user, and synchronization traffic on that link, plus a management plane, precisely enough that in principle a radio from one supplier and baseband from another can talk on the same site.

That alone would make this a real week. But the architecture the alliance published does something I was not expecting. It does not just split the base station into pieces, it adds a brain. There is now a RAN Intelligent Controller in the design, a piece of software whose entire job is to watch the radio network and tune it in near real time. Third parties are meant to write apps that run on it, reaching into the network over a new interface to change how the radios behave. A separate, slower controller sits up in the management layer for the decisions that can take their time.

I keep circling that controller because it changes the story. Every earlier chapter here was about breaking the box open and mixing vendors, a supply-chain argument dressed up as architecture. The RIC is a different ambition. It treats the radio network as a programmable platform, something you optimize with software you did not have to buy from the radio maker. If that lands, the interesting money in telecom stops being the metal on the tower and starts being the apps deciding what the metal does.

My skepticism has not gone anywhere. A specification precise on paper is not the same as two vendors interoperating on a live site under real traffic, and the alliance knows it, which is why plugfests are coming this year to try to prove it. Fronthaul timing is unforgiving, the synchronization budget is measured in fractions of a microsecond, and there is a long history of interop that works in a lab and falls apart in the field. The RIC has an even harder road, because writing an app that safely retunes a production radio network is not a hackathon project.

Last year I asked whether they meant it. This year they handed me a document precise enough to argue with, which is more than most telecom alliances ever manage. The fronthaul spec is the boring, decisive part. The controller is the wild card, and it is the thing I will be watching to see whether anyone builds an app for it that a carrier actually trusts on a live network.

Sources