[Cryptography] Proper Entropy Source

[Patrick Chkoreff]

John Denker via cryptography wrote on 1/22/20 6:53 AM:

> A PRNG depends on computational complexity, and also
> depends on a seed.  We still need a good RNG to produce
> the seed.

I'm not entirely clear why seeding is portrayed as so difficult.  Please
bear with me here.

I just rolled a sixteen-sided die to get a "truly random" hexadecimal
digit, which I'll call $digit here so as not to disclose this vital
secret.  Then I ran:

$ echo $digit >/dev/random

I'm pretty sure I just made the state of my PRNG 16 times more difficult
to guess.  I could continue with more dice rolls to make it fiendishly
difficult.

Is the alleged difficulty of seeding a PRNG based on the problem of how
to do it automatically?  If so, why is that difficult?  Isn't the
sequence of key strokes, mouse movements, internal interrupts, etc.
enough to inject sufficient entropy?

Or, is the alleged difficulty of seeding a PRNG mostly confined to the
problem of how to do it on a fresh machine upon first starting up --
e.g. spinning up a new virtual machine?  And why is so much importance
attached to the need to generate ssh keys immediately upon spinning up a
brand new machine?  I guess I'm not in the business of running large
server farms.  Is it too much to ask to seed a new machine somewhat
manually from some outside source when it first starts up?


-- Patrick