- every non-trivial software is a compiler
- every non-trivial software is a database
Every non trivial software have to read/parse/manipulate/store a lot of data. I think it ultimately permit to simplify thinks to just admit that we are doing a specialized database/compiler. The fact that input data are coming from file, database, network, command line or the code itself and are going to files, database or network is just an implementation detail.
The solution is to design your software as a compiler from the start.
Too bad. It's an interesting talk on the architecture of an ambitious compiler redesign. The title is a simplification for the sake of being a keynote.
There was a lot of interesting comments yesterday on a post called "Presumption of stupidity" (https://news.ycombinator.com/item?id=10034883). This could also applied to comments: Do not assume the worse you can imagine by default.
And yes, it's indeed a very interesting talk.
Actually this is even more true than you probably think it is.