Since you are talking about something that is almost scientific use scientific notation NOT the funny US monetary notation in which a billion is only a thousand million and not a million million.

400 billion of which nobody knows whether they're actually needed or not, but they won't delete them "just in case".

After all, disk is cheap (pound in swearbox).........

"...commissioning a report on the effects of de-duplicating the digital universe. Preliminary research indicates that it could shrink by up to 70 per cent,..."

Anyone who has ever worked with large databases *knows* that there is not such thing as an ONF database. For those of you who don't grock, ONF means that the database has no data-replication *anywhere* except for keys (duh).

While ONF is a worthy goal on paper, real life HW constraints means that to get data in decent time, *some* data replication is required. The actual amount of replication varies from DB to DB depending on the needs of the business, but it is *always* there.

Same thing here - some "information" is so retrieval-time-critical that local duplication will be preferable to singularity.

And five minutes after the de-duplication is completed, a hardware failure occurs and all their nicely redundant data backups are no longer there...