Meso-computing and meso-data: the forgotten middle
milliams.com
milliams.com
Now in computing I do the same. I don’t do fancy buzzword big data and my cloud experience is single instances. But the problems I solve are hard and there’s no magic recipe I can steal from SO to solve them. It’s all novel problem solving.
The money is in the meso scale. Small stuff isn’t interesting and the cloud providers have taken the margin out of the big stuff (or at least have taken the margin for themselves).
It’s also fun to solve these novel problems...though it does make me a bit of a dinosaur and I can’t complete at getting a job at a FAANG because I don’t do stuff “at scale”.
Besides the meso-scale identified by the author of the linked article, I think there's another little-explored island at the scale where programs are bigger than what we built for 16-bit machines, but smaller than what we tended to build for 32-bit machines. Certainly most extended memory DOS games and early Windows programs could be argued to lie in this range, but I'm asking what would we build in this scale using modern tools for contemporary applications?
For example a GUI program in the 50Kb - 500Kb scale would probably be written in C/C++ and use #defines for Windows message constants. A program in the 10Mb - 100Mb scale might be written in Java or C# and include a 10's to 100's of megabyte runtime framework. (I'm including the framework, not just the user code, in the scale here.) What about the scale from 500Kb - 10Mb? I could imagine a program, still written in C/C++, using Sqlite (which didn't exist 21 years ago) to define its Windows message constants.