Hector Martin

64bit internal, 53bit external... this has been known for like 20yr

here, 6yr ago... http://2ality.com/2012/07/large-integers.html …

You do not understand floating point numbers. 2^53 is not the largest integer that can be represented accurately in 64-bit floats. *All* the integers *from zero to 2^53* can be represented, but *many more larger than that* can too. Just not all of them.

just read the URL i linked, it explains things very well

or... i dunno, argue with brendanhttps://twitter.com/brendaneich/status/526826278377099264?lang=en …

Yes, the URL you linked explains it: beyond 53 bits, you can only represent every 2 integers (and then every 4, then every 8, etc). But *you can still represent those integers*. The integer I'm using as an example is one such integer you *can* represent. Read it yourself again.

yeah, you can, and you can also get rounding errors based entirely on the engine... you can't do that in production, so most don't consider it an option

No, you cannot get "rounding errors based on the engine". IEEE 754 is *precisely* defined. The *same* integers are, or aren't, representable in *every single system using IEEE 754". The only problem here is toString rounds some integers to impossible ones for no reason.