I Can't Believe It's Not Zero!

Proposal: people prefer “mantissa” because it abbreviates better than “significand” (any truncation of which collides with “sign”). But “mantissa” is formally incorrect, since it already has a different meaning. So we should use “magnificand” instead.

- code that uses (s,e,m) still makes sense
- it’s the thing that is magnified by the exponent—the magnificand
- magnificent

Feb 12, 2023 at 1:45pmWeb
Show threadMiguel de Icaza ᯅ🍉Feb 12, 2023
@steve I endorse this cultural change
Show threadBen LangmuirFeb 12, 2023
@steve more like (s, e, 😍)
Show threadYusef NaporaFeb 12, 2023
@steve my only objection is that mantissa make me think of manticores, which are cool. But magnificand is très magnifique
Show threadTom ForsythFeb 12, 2023
@steve This is the sort of pragmatic pedantry that gets me regularly burned at the stake too. I got your back, brother.
Show threadDaniel HolthFeb 12, 2023
@steve mantissa is a magnificent mother mantis
Show threadaegilopsFeb 12, 2023

@steve that would "correct" zero real-world issues.

I'd never heard the term "mantissa" before seeing its use in FP - I've never needed to use a logarithm lookup table! They are obsolete.

Terms in mathematics and computer science are often appropriated by analogy, and that's all that happened here. It seems like you object to this?

Please don't spend effort on something that "solves" a non-problem and would lead to more confusion.

#mantissa #FloatingPoint #pedantry

Show threadDaniel McLauryOct 30, 2023
@aegilops @steve Then you don't work in an appropriate context to offer an opinion here. This is obviously targeted towards people who have to use this sort of terminology in their day-to-day work, and there are a great many of us.
Show threadaegilopsOct 30, 2023

@danielmclaury

Changing terms in computer science isn’t worth it to satisfy a feeling of discomfort.

I reject your attempt at gatekeeping - no thanks to _that_ attitude!

Show threadDaniel McLauryOct 30, 2023
@aegilops Figuring out exactly what to name stuff is actually worth spending more than 50% of your time on, because the payoff to it is far more than a 2x multiplier to productivity. Does Microsoft not spend a lot of time inculcating this attitude in their people any more? It has famously been one of the main things they tended to get right.
Show threadCassandrichFeb 12, 2023
@steve I'm perfectly fine with just redefining mantissa to match actual usage. Call me a heretic. It's far less offensive than redefining crypto. 🤷
Show threadI Can't Believe It's Not Zero!Feb 12, 2023
@dalias I mean, sure, but magnificand is actually a better word for it anyway, so
Show threadCassandrichFeb 12, 2023
@steve I'm confused. To me, magnificand would specify the magnitude, i.e. denote the exponent.
Show threadFCLCFeb 13, 2023
@steve Honestly unsure if this is real or a shitpost, and I don't know which I prefer anyway
Show threadI Can't Believe It's Not Zero!Feb 13, 2023
@fclc I’m not sure either!
Show threadI Can't Believe It's Not Zero!Feb 13, 2023
@fclc honestly most of my best ideas started life as elaborate trolls, so…
Show threadChrister EricsonFeb 13, 2023
@steve except that’s the “fraction”. Mantissa is the fraction together with the implicit 1-bit.
Show threadTiago PeixotoFeb 13, 2023
@steve What is the different meaning of “mantissa”?
Show threadI Can't Believe It's Not Zero!Feb 13, 2023
@tiago traditionally (pre-floating-point), “mantissa” meant the fractional part of the logarithm, used in log tables for computation. E.g. using the base-10 log of 150 as an example, the “characteristic” (integer part) is 2, and the “mantissa” (fractional part) is something like 0.1761.
Show threadI Can't Believe It's Not Zero!Feb 13, 2023
@tiago that’s closely related to, but not the same as the thing we call “significand” or “mantissa” when we talk about floating-point (in the example above, the significand would be 10^0.1761 = 1.5 if we normalized decimal FP like we do binary FP).
Show threadTiago PeixotoFeb 13, 2023
@steve Understood! I've never seen this original meaning of mantissa before. It seems its usage has died out together with logarithm tables.
Show threadTaylor R CampbellFeb 13, 2023

@steve This would also inspire endless insufferable debates about whether the ⟨c⟩ is pronounced /k/ (as in significand) or /s/ (as in magnificent), like the ⟨g⟩ in gif or the ⟨a⟩ in data.

Whether that's a pro or a con is left as an exercise for the reader—insufferable as those debates may be, they would serve to spread the word, so to speak.

Show threadTaylor R CampbellFeb 13, 2023

@steve I would like to point out that I personally used the word mantissa correctly just four months ago, when demonstrating, with a worked example from a printed book of tables, why a table of haversines works better than just punching 1 - cos(t) into a computer for navigating a clipper in 1845.

(Admittedly, I didn't _write_ the word mantissa; I just _read_ it from the book while working the example, but that's still usage of the word, just in a different I/O direction.)

Show threadTaylor R CampbellFeb 13, 2023

@steve Question: For binary floating-point, in 1.101₂ × 2³, is the magnificand 1.101₂ (i.e., 1.625), or is it 1101₂, or is it 101₂?

For decimal floating-point, in 1.23 × 10⁴⁵, is it 1.23, or 123, or 23? What about 2.34 × 10⁶?

And for subnormals?

Or will it be just as ambiguous about this as significand/mantissa, for fun?