3 ms·
You're vastly oversimplifying. In physics and finance, the "initial time" has always been t_0, not t_1. This has absolutely nothing to do with machine code, or
by goodside 15y ago
You're vastly oversimplifying. In physics and finance, the "initial time" has always been t_0, not t_1. This has absolutely nothing to do with machine code, or with "blind cargo-cult copying". There are intrinsic reasons for doing it this way, which I'll leave for you to ponder yourself, ideally in a quiet room in a lotus position.
Traditions are not always arbitrary. Don't dismiss them until you're damned sure why they've survived, especially when you know for a fact that quantitative fields filled with high-order geniuses haven't bothered to scrap them. "Cargo-cult" thinking is a pernicious influence, but it's exceptionally rare that it's the complete explanation of anything.
- edtechdev 15y agoI'm addressing the topic of the post: the use of indexes in programming, not names used for variables that store an initial time. myarray[2] doesn't mean 'myarray at time 2'
- mjw 15y agoOne advantage I think the parent is alluding to, is that zero-based indexing makes it a bit easier and more natural to generalise between continuous (functions on R) and discontinuous (functions on N) situations. Since the latter are often used to approximate the former, this is quite handy.