Prof. Anima Anandkumar

1) I recently visited a "big" high energy accelerator (don't want to reveal since I actually admire people working there). I was taken aback at shoddy data preservation practices. It is astonishing that data from experiments that cost billions of dollars are never logged

I terms of data archiving, I am not aware of a single lumiblock lost so far. Where we struggle, is to move the SW to process it 20yrs along with technology changes. We can always process past data with past SW, but maintaining every branch of tens of millions lines of code costs.

I did not want to name the lab, so practices can differ from what you have seen. But even at @CERN I thought data is preprocessed before storing. The problem is this processing is not invertible. And scientifically not the optimal thing to do

We write raw data (i.e. unprocessed bytestream) from the detector directly onto tape for permanent storage. No loss there.