I work in a hospital and alot of the things he talks about would be quite cool.
Immutability: we have one big as table with tons of stuff in it (this is not the db design you learned in school) and and even bigger table of all the changes made. The way this is handeld in datomic would be optimal.
Querys: Query all the data based on a time stamp would be really useful. Querys where I could combine the DB data with some other data I got from other applications/databases, heap or somewhere else.
I never understood Amazon's stance. Their whitepaper and case studies just says medical companies have built applications on our cloud. What they fail to mention is whether they were only able to do so by anonymizing the data prior to upload which means in fact there isn't any PHI on their clould.
They seem to circumvent these disclosures for the obvious reasons of liability and the fact that today you really need a signed Business Associate (BA) agreement with any 3rd party company that may host PHI. Hospital compliance officers would have a heart attack if they hear 'clould' and no BA offered.
The whole AWS medical position feels very hand wavy--yet another reasons this whole medical industry has such a hard time innovating.
I worked for a company that had a large amount of HIPAA data on Amazon's EC2 platform. We did significant encryption (including field-level encryption) on the data to help mitigate security concerns.
We were able to pass security inspections by 4 fortune 500 health companies with this model (though a couple of them needed significant hand-holding).
It made querying that data a pain in the ass because you couldn't do a SELECT ... WHERE <encrypted field> = 'foo' as the data wasn't encrypted client side.
One cool thing about datomic is that the query engine runs client side, so it can also invoke client side methods as part of the query. This means that if you had encrypted data in your store, that it could be trivially decrypted client-side and you could potentially get the full expressiveness of the language to work with your data.
I don't know what legal requirements we have, but I don't think EC2 would work, atm its all on servers inhause. I was talking about the concepts and ideas not the way datomic implements them atm.
Immutability: we have one big as table with tons of stuff in it (this is not the db design you learned in school) and and even bigger table of all the changes made. The way this is handeld in datomic would be optimal.
Querys: Query all the data based on a time stamp would be really useful. Querys where I could combine the DB data with some other data I got from other applications/databases, heap or somewhere else.