On Tue, Dec 22, 2020 at 7:30 PM Keith Lofstrom <[log in to unmask]> wrote:
>
> This isn't Scientific Linux, though SL and the team that
> supports it would be involved in implementation.
>
> ----
>
> AFAIK, we are still in the middle of Long Shutdown 2,
> with the Large Hadron Collider /not/ sending terabytes
> of experimental data through the dedicated HEP network.
>
> Meanwhile, gathering musicians on stage for a symphony
> orchestra is a big health risk during the COVID pandemic.
>
> The speed of sound across a symphony stage is 500,000 times
> slower than bits on an optical fiber.  In theory, musicians
> could be connected through the HEP network, spread out over
> 5000 kilometer distances compared to the 10 meter distance
> across a symphony stage.  Distances are smaller than that
> between groups of European or North American cities.
>
> There are probably more symphony fans than high energy
> physics fans (many physicists are both), so using the HEP
> network for concerts during the COVID crisis could earn
> a LOT of political capital, and help with future funding,
> including funding for the next upgraded HEP network.
>
> It would also develop new techniques for synchronizing
> planet-scale sensor networks.  There are likely some
> excellent astronomical and geophysical uses for that.
>
> I'd guess that readers of this list know the people who
> know the people who know how to do this.  What could we
> slap together in a hurry?
>
> Keith

The HEP network, I don't know. I used to work in high-frequency
trading, but that was.... very asymmetric traffic, and merging it
wasn't the point. Being *first* to place an order was the point. I got
clear of the business just in time, when FPGA's parked on the fiber
optic coming out of the stock exchanges replaced the very, very
expensive low-latency multicast channels running cross-country to
remote data centers. You don't *want* distributed signals for that,
only your personal signal mattered. You *wanted* everyone else to get
signals slightly later, and get your responses in first. The potential
for phase delays to cause positive feedback loops scared the *hell*
out of me when I talked to the math or stock parts of the company.

I think you have a *lot* of fundamental problems to deal with. One is
that even so-called "real-time kernels" and "real-time" systems often
are not, they merely flatten the delay for consistency by smoothing
out the interrupts common to CPU based signal processing. It looks
great on an oscilloscope after this, but it's smoothing out "sharp"
sounds, especially when also playing with automatic gain control and
usually trying to synthesize what the designer thinks should be the
information that matters, rather than presenting as much of the
original data as possible to the only part of the system that matters:
the human nervous system. The results are.... well, they look great in
a PowerPoint slide but they tend to not actually provide intelligible
sound.

Did I ever mention I designed and built the first *stereo* cochlear
implant stimulators? A very few people had two stimulators back then,
but they were always distinct models of stimulator and impossible to
cross-wire for stereo sound until I built a rig.