Blanking Out | Hacker News Reader

Blanking Out

sarahwerner.substack.com

48 pointsgruseom7 comments

There is a similar problem with digitized photos, for instance from glass plates. Often they had notes in margins or on the backside, but many scanning efforts didn't take these into account. So a lot is lost forever, either because the originals are destroyed, or forgotten in some archive.

In the precomputing era you could (legitimately!) get a PhD for building a table of constants (e.g. log tables early on, later CRC data like steam tables and curves). In the humanities you could get one for writing a concordance of all the words in Shakespear or some other corpus, or all metaphors, and the like. It's obvious how computation killed them.

But in a few years some enterprising grad student or post doc will earn accolades by re-scanning some old sources, taking into account some important metadata which right now is irrelevant. Perhaps labeling or annotating the result in some novel way.

> But in a few years some enterprising grad student or post doc will earn accolades

As they should! These sort of things are properties of the knowledge building processes themselves, and that is precisely one of the goals of research work - to uncover insights in places where people before you thought there were none.

There's an interesting article somewhere that I really wish I could find right now - the gist is that someone working at I think a dictionary publisher was confused to find a filing cabinet that contained an index of every word reversed in alphabetical order. It took some thinking to realize that this was a useful index to find all of the words that ended in a given suffix - e.g. go to "gni" to find the gerunds.

It's interesting to think though that maintaining a dictionary once meant having a large set of filing cabinets containing sheets of words sliced and diced in every way you might need.

I've been saying this for years! There's a lot of hype in tech around "backing up" physical things with scanning technologies. But you simply can't do it perfectly. There's no such thing as a perfect "backup" because reality is aggressively analog.

A digital copy of a painting is way better than no copy at all, but it's not a replacement. You can't carbon date a digital scan, etc.

I don't remember where I heard this but: digitization is the art of deciding what to forget

A 2009 paper shows that you can tell blank pages apart from one another by examining them closely under a scanner: https://citp.princeton.edu/our-work/paper/

So not only is there so much texture there, but even a single manufacturer's individual blank pages differ dramatically because of physical wood fibers within the page.

Reply on news.ycombinator.com