Not directly IT related, but involves SSL certs and IoT:
A place I knew (involved in IoT roughly a decade ago) had written its own SSL certification validation logic (not the crypto part, mind you: just the part that checks “is the date on the cert still valid now?”) for the very first version of its hardware.
It was rolled out in the summer and things went fine. Or so they thought.
On Jan. 1st panic broke out, as the previously undetected off-by-one programming error in the validation logic meant devices could not validate the server’s cert anymore.
There were quite a few layers of learning in that fuck-up, but thankfully it was solved within a few days :)
Beyond the obvious "IT needs will likely still be there", I have two possibly contradicting hunches:
- on one hand, can software in general be a definite "plus" on the question? Should we maybe look for solutions that don't involve too much software, as the footprint of everything software (taken widely and including intensive machine learning, for example) is already pretty big?
- on the other hand: assuming software is required "in large amounts", maybe there is room for heavily optimising all the things involved in building and running software.
With the observations above, if I'm interested in continuing to work with software, working on computational efficiency questions could be interesting and possibly useful.
Edited to add: it’s the graphviz generator -> https://thrift.apache.org/tutorial/graphviz.html
I’d guess gRPC and other such IDLs have a way to render similar things as well.