Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Last POWER1 CPU on Mars Is Dead

talospace.com

Post reply on HN
Print view View original

91–100 of 145 posts

Re: The Last POWER1 CPU on Mars Is Dead

#91

Post by tomp »

gmueckl wrote: post #45
tomp wrote: post #23

Earlier quoted context omitted.

Really? Can't you just anchor to e.g. one revolution of the planet? That should be synchronizable...

No. You would have to observe the planet from your position and that observation changes based on how you move relative to the planet. The whole of SRT and GRT work in such a way that no general clock synchronization can ever exist for observers in different frames of reference. If you take a reference with you on a journey through the solar system, that reference will stay accurate for an observer travelling with it…

What I had in mind is, you can observe your planet's position and velocity (relative e.g. to center/rotation of the galaxy), and the target planet's relative position and velocity, and use both to translate the observations into virtual "stationary" observations... And "unit of time" would be defined as 1/n of one planetary/galactic revolution, so every observer would be able to calculate their own "galactic time" independently. Well, maybe GPS already does something similar.

Re: The Last POWER1 CPU on Mars Is Dead

#92

Post by icebraining »

tomp wrote: post #27
dahfizz wrote: post #8

Earlier quoted context omitted.

I don't really see the problem with standardizing on 'seconds since the epoch' for any interplanetary communication and having clients convert to something more meaningful to humans. Any other "universal" time format will be just as meaningless when days and seasons are all different lengths.

Yeah I don't entirely understand why we don't that already... Like, websites should be sending UNIX time to clients, and then the browser should display it in users local timezone/format.

HTML already has a tag, which one can use to send a machine readable representation of the textual content; it would be easy for browsers to display it in local tz/format. Of course, almost nobody uses it, as usual.

https://www.w3.org/TR/html52/textlevel-semantics.html#the-ti...

Re: The Last POWER1 CPU on Mars Is Dead

#93

Post by tomp »

hermitdev wrote: post #46
tomp wrote: post #23

Earlier quoted context omitted.

Really? Can't you just anchor to e.g. one revolution of the planet? That should be synchronizable...

No, you can't. Atomic clocks measure the oscillation of atoms (I think Cesium is currently used), and though constant at a fixed point, vary when theyre moving at different speeds. GPS wouldn't be possible without taking this into account, for example. A second on earth will never be the same as a second on Mars.

A second is duration, but we're really interested in instants. Why can't you can define an instant independently of the observers, e.g. in reference to the rotation of the galaxy (i.e. "this happened at angle 37 from Milky Way - Andromeda line, at distance 403 parsecs from Milky Way center, at 692.3 Milky Way rotations since epoch (Big Bang)").

Re: The Last POWER1 CPU on Mars Is Dead

#94

Post by oso2k »

jakeinspace wrote: post #87
oso2k wrote: post #68

Earlier quoted context omitted.

There are real, standardized protocols [0] for this sort of thing, now. Early in my career, when I worked at JPL/NASA, I did a bit of work to test these protocols on Linux in a simulated Network of spacecraft. Fun stuff! [0] https://en.wikipedia.org/wiki/CCSDS_File_Delivery_Protocol

What a coincidence, I'm working on the core flight system project right now, which includes an implementation of CCSDS File Delivery Protocol. You might be happy to know that project is still very much alive within NASA (you were probably already aware).

Haha. Great to hear that work still lives! I hope you’re not using the (horrible in retrospect) stuff I wrote 15 years ago that worked on 2.0-2.6 kernels. I remember porting the network kernel driver from 2.0-2.2 manual build system to implementing a (really basic) Makefile that worked with 2.0-2.6.0ish kernels (which were newer at the time). I knew Scott Burleigh was still innovating CCSDS and other protocols at JPL, and, IIRC the work I had done had also been delivered to Ames and ESA for test lab use. I had some other work to create protocol dissectors for Ethereal (before the name change to WireShark). WireShark now includes built in dissectors for CCSDS, CFDP and others.

Re: The Last POWER1 CPU on Mars Is Dead

#95

Post by oso2k »

gcb0 wrote: post #78
oso2k wrote: post #68

Earlier quoted context omitted.

There are real, standardized protocols [0] for this sort of thing, now. Early in my career, when I worked at JPL/NASA, I did a bit of work to test these protocols on Linux in a simulated Network of spacecraft. Fun stuff! [0] https://en.wikipedia.org/wiki/CCSDS_File_Delivery_Protocol

What does it use for time? The spec mentions time in a lot of places, but never defines it. There is even a field to communicate light-time but no mention of how to synchronize anything. I guess this runs on top of another protocol that solves it? Any idea what it is?

So, it’s been a while since I’ve dived in [0], but “time” as a dependency has been, what I would call, “engineered out of the system”. Instead, there’s concepts of (local) timers, Store-And-Forward Overlays, retries, communication opportunities and the overall realization that time is relative and synchronized time can be difficult at a protocol level when link-level distances are always changing (by miles to billions of miles at a time per communication opportunity).

[0] https://public.ccsds.org/Pubs/720x1g3.pdf

Re: The Last POWER1 CPU on Mars Is Dead

#98

Post by hopler »

CydeWeys wrote: post #65
wlesieutre wrote: post #18

Earlier quoted context omitted.

At a certain point of accuracy you run into issues with this - seconds don't pass at the same rate due to relativistic effects (both speed and gravitational). An example we already deal with is GPS satellites orbiting Earth. Their clocks tick a little bit slower on account of how fast they're orbiting, and tick a bit faster on account of being further up Earth's gravity well. The gravitational effect is stronger, and…

GPS satellites already control for relativistic effects. Their required accuracy is so high that they wouldn't work without doing so. You would simply also correct for relativistic effects caused by the orbits of different planets, and boom, you have a pretty consistently defined time that is valid across the entire solar system.

Consistently defined time is less useful when events at time 12:04 cause events at time 12:02

Re: The Last POWER1 CPU on Mars Is Dead

#99

Post by hopler »

tomp wrote: post #93
hermitdev wrote: post #46

Earlier quoted context omitted.

No, you can't. Atomic clocks measure the oscillation of atoms (I think Cesium is currently used), and though constant at a fixed point, vary when theyre moving at different speeds. GPS wouldn't be possible without taking this into account, for example. A second on earth will never be the same as a second on Mars.

A second is duration , but we're really interested in instants . Why can't you can define an instant independently of the observers, e.g. in reference to the rotation of the galaxy (i.e. "this happened at angle 37 from Milky Way - Andromeda line, at distance 403 parsecs from Milky Way center, at 692.3 Milky Way rotations since epoch (Big Bang)").

Distance depends on how fast you are moving. As do the endpoints of a rotation.

Re: The Last POWER1 CPU on Mars Is Dead

#100

Post by ajdlinux »

inamberclad wrote: post #97

Do people have a particular attachment to the power architecture?

There's still a reasonable PPC Mac/Amiga hobbyist community, many of whom will tell you about how much purer PowerPC is compared to the evil ugly x86. As someone who does PPC Linux kernel work the pragmatic part of me would like to drop support for a lot of these older machines, but as long as the machines are still actively used the support gets maintained.
Post reply on HN