Storage Capacity

[Isolder74]

BTW gultural knowlrdge = Cultural Knowledge

Sorry

[Darth Servo]

Modern computers already utilize various forms of streaming data processing techniques; there is no reason whatsoever to assume that this technique has been lost to 24th century engineers.

I'm not so sure. Consider how many of their own technological tricks they forget about.

And of the various cells in your body, they all have identical DNA, for example, so there's no need to store more than one copy of that.

Not quite. The DNA of the white blood cells gets modified to some minor extent.

[SWPIGWANG]

You are, of course, assuming that every language can be stored in the form of combinations of a small number of letters. Given that many of the languages known to C3PO have no written form at all, some cannot be pronounced by humans at all, and others may be even more esoteric, I'd say you're grossly underestimating the difficulty.

The language does not require written form or to be humanly pronouncible to be stored. As long as there is an way of codifying the language into an simpler form the information can be compressed greatly. After all, an computer knows languages not though voices or words, but good old numbers. Unless an language is effectively patternless (which is unlikely because such an language would be utterly inefficient), there is always an effective way of compressing.

Also there is an limitation to how esoteric an language can be that is bounded by the memory capacity of the biological species that created them. While known biological brain is capable of huge amount of storage (or huge amount of distortive compression, actually) the amount where an average mind can remember sufficiently clearly for the purpose of communication is highly finite. Human beings for example, are limited to merely a few thousand words of effective vocabury. Considering the fact that humans dominate the StarWars galaxy, there are probably few species capable of outsmarting an human or have languages magnitudes more complex.

Of course, I would NOT want to translate languages used by culture minds, but this is not culture mind's language we are talking about.

With 400 million stars and the need to know where all of them are at a bare minimum in conjunction with last known velocity data on all of them and astrophysical projections of all their movements to sufficient accuracy to drop a ship within a few tens of thousands of kilometres of the target planet anywhere in the galaxy at any time, it's a pretty stressful computer exercise to say the least.

We don't know how stars effects hyperspace and we don't know how accurate the calculations have to be. Also, while calculating the said astro-projection is difficult (and impossible to get perfect accuracy thanks to nonlinear parital differential equation grrrr) it is not difficult to store an data table for projections within a reasonable timeframe. Assuming planck's distance accuarcy, ~180 bits or 30 bytes is required for each position vector. Now given the accuacy of say one second with storage for an week and 4x10^10 objects to keep track of. We would only have 1e18 bytes of data. This translates to 1,000 PB of memory. Not an small value, but not completely uber as far as scifi goes.

[General Zod]

We don't know how stars effects hyperspace and we don't know how accurate the calculations have to be. Also, while calculating the said astro-projection is difficult (and impossible to get perfect accuracy thanks to nonlinear parital differential equation grrrr) it is not difficult to store an data table for projections within a reasonable timeframe. Assuming planck's distance accuarcy, ~180 bits or 30 bytes is required for each position vector. Now given the accuacy of say one second with storage for an week and 4x10^10 objects to keep track of. We would only have 1e18 bytes of data. This translates to 1,000 PB of memory. Not an small value, but not completely uber as far as scifi goes.

actually. according to various TMs that i've read, when making hyperspace calculations everything involved is taken into account. departure space, destination, and the gravity wells of major astronomical bodies. since the galaxy doesn't stay still and is constantly moving, even if very minorly, every fluctuation of gravity has to be taken into account for the hyperspace jump so they aren't thrown off course completely and go headlong into a star or an asteroid.

there are a few safely charted hyperspace lanes, but these are only the most well traveled paths and mean that they don't have to perform nearly as many calculations.

[SWPIGWANG]

and the gravity wells of major astronomical bodies. since the galaxy doesn't stay still and is constantly moving, even if very minorly, every fluctuation of gravity has to be taken into account for the hyperspace jump so they aren't thrown off course completely and go headlong into a star or an asteroid.

If the drone does not have to keep track of astronomical bodies by itself, than it can easily do the calculation for gravitational fields the spacecraft would go though. It is simply addition, and adding say 400million numbers is not exactly hard, even if one needs to do it iteratively many many times to ge the precise result. (depends on the increments used for time) Unless we know how accurate the said calculation is required, I can claim that my PC can do it.

Besides, gravity decays at the rate of d^2 and much of the effects can simply be clipped out if required. Starwars hyperdrives have limited course changing capacity within hyperspace, which is probably sufficient to clear up any error from starsystems a few dozen light years away. (unless of course, gravity do really weird things in hyperspace with crazy relativity....)

[Mad]

400 million stars is an estimate for the number of stars that a hyperspace calculation would have to deal with, right? (As in the rest are ignored in a typical case for being too far away to worry about once the general direction is known.) Or is there some source that gives the GFFA far less stars than the Milky Way's 100 billion?

Storage would require at least the 100 billion stars to be held. We know that some ships, like X-wings, can skim by close enough to stars that a slower ship would be pulled out of hyperspace by the gravity well of said star (course plotting stuff from X-Wing novels). We also know that slower ships exit hyperspace near planets. Which means, at the very least, that an X-wing can hyperspace right through star systems. Which means each star's satellites must be stored, too. We're getting well beyond 4E10 objects now.

What's the complexity of a "find the straightest course that doesn't hit anything" algorithm? I haven't found anything applicable, just routing algorithms (would work for multiple jumps, but that's Rebel strategy; those who don't need to hide their location can plot a curved course and use much less fuel) and on-the-fly computations for robotics (which won't work at all, especially with the 90 degree turn things going on).

It'd could be along the lines of taking a line and then generating some kind of sine wave function and modifying it for each obsticle it finds to get the smoothest possible course. After each modification, it'd have to check all relevant stars to make sure there's no collision with too strong of a gravity well. If it enters a star system, it'll have to check against all objects in that system. The amount of time spent calculating would vary based on how many retries are required to generate a good path. With 400 million stars and the number of objects for in-system skims, there could potentially be a lot of re-checking.

[General Zod]

What's the complexity of a "find the straightest course that doesn't hit anything" algorithm? I haven't found anything applicable, just routing algorithms (would work for multiple jumps, but that's Rebel strategy; those who don't need to hide their location can plot a curved course and use much less fuel) and on-the-fly computations for robotics (which won't work at all, especially with the 90 degree turn things going on).

things don't stay still in interstellar vacuums. you have to compensate for the movement of known objects when making hyperspace jumps as well or you risk splattering right into one.

[Mad]

things don't stay still in interstellar vacuums. you have to compensate for the movement of known objects when making hyperspace jumps as well or you risk splattering right into one.

With the proper orbital information plugged in, the computer can figure out where anything will be at any given time. Since the ship knows where it should be at any given time after the jump is made, matching up ship.position.at(time) and star.position.at(time) isn't difficult. It won't add to the computational complexity I was asking about.

Now, if the computer was calculating all the orbital information on the fly, then the computation complexity will jump to O(n^2), and it'd have to account for all the mass... With 100 billion stars, that means that if the computer in ANH took as long as it did to crunch the numbers, then if that same computer were to use precomputed orbital information, it'd be able to compute it almost instantly.

With that kind of performance increase, I can't see why they't use real-time when a precomputed optimization will have the same effective results and be much, much quicker. (Both methods will require periodic updates to the star charts for changes that happen for various reasons, such as superweapons or new sattelites brought into service, so inaccuracies won't be able to propogate very far either way.)

[PainRack]

The language does not require written form or to be humanly pronouncible to be stored. As long as there is an way of codifying the language into an simpler form the information can be compressed greatly. After all, an computer knows languages not though voices or words, but good old numbers. Unless an language is effectively patternless (which is unlikely because such an language would be utterly inefficient), there is always an effective way of compressing.

A computer is unable to speak the language fluently. May I remind you of
"All your base are belong to us!"

Also there is an limitation to how esoteric an language can be that is bounded by the memory capacity of the biological species that created them. While known biological brain is capable of huge amount of storage (or huge amount of distortive compression, actually) the amount where an average mind can remember sufficiently clearly for the purpose of communication is highly finite. Human beings for example, are limited to merely a few thousand words of effective vocabury. Considering the fact that humans dominate the StarWars galaxy, there are probably few species capable of outsmarting an human or have languages magnitudes more complex.

Of course, I would NOT want to translate languages used by culture minds, but this is not culture mind's language we are talking about.

May I remind you again of ancient times, when story-tellers would memorise entire epics by heart? An extremely impressive feat if one memorises the Indian epics, which has stories lasting over 3 generations long. And they routinely memorise several of them, including religious books like the Vedes and so forth.

We don't know how stars effects hyperspace and we don't know how accurate the calculations have to be. Also, while calculating the said astro-projection is difficult (and impossible to get perfect accuracy thanks to nonlinear parital differential equation grrrr) it is not difficult to store an data table for projections within a reasonable timeframe. Assuming planck's distance accuarcy, ~180 bits or 30 bytes is required for each position vector. Now given the accuacy of say one second with storage for an week and 4x10^10 objects to keep track of. We would only have 1e18 bytes of data. This translates to 1,000 PB of memory. Not an small value, but not completely uber as far as scifi goes.

We do know that even stray comets(Dark force rising) and asteroids( Rebel Dawn) can affect hyperspace jumps for smaller vessels. Han comments about supernova and the like represent an alternate end of the spectrum. To put it simply, we have to use super-computers now just to safely send our rocket ships in our own system. Imagine the drastically larger scale needed, to move ships on an inter-stellar basis.

[SWPIGWANG]

May I remind you again of ancient times, when story-tellers would memorise entire epics by heart? An extremely impressive feat if one memorises the Indian epics, which has stories lasting over 3 generations long. And they routinely memorise several of them, including religious books like the Vedes and so forth.

Which is also, well, nothing in comparison to my harddrive.

Take for example, the Iliad....w00t it is 757kilobytes (translated), OMG thats alot of memory. (not)
http://www.gutenberg.net/etext/3059

I can only store only 10,000 or so stories of that size, because my computer suxx0r. (I need an upgrade damn it, can't even run CMBB smoothly at high point level for god sakes....)

A computer is unable to speak the language fluently. May I remind you of
"All your base are belong to us!"

Storage capacity is not intelligence or processing capacity. "All your base" was a sentence written by humans, by the way. An computer can easily be programmed to not make mistake like that.

Just consider that the computer in front of you can easily store hundreds of encyclopedia worth of text.

Lets consider english, an extreamly diverse language by earth standards.
http://hypertextbook.com/facts/2001/JohnnyLing.shtml

So we get 1e6 words roughly. (! thats alot actually) Now we say we use 1,000 additional words to cross reference definitions, than we have 1e9 words. Assuming each word to be 10 letters long, we get 1e10 bytes of data, or 10GB as an maximum required. (before compression, which can reduce the amount need by at least an magnitude or two.

As for an book, we can calculate the required memory by word count. Assming 200 pages and 500 words per page, we get 1e5 words and 1e6 bytes per book. At 10GB we can store 1000 books, more what an average english speaker will read in an lifetime. (thats ~50 books/yea or 1 in 10 days or so) (reference: http://www.writersservices.com/wps/p_word_count.htm)

That is more than enough information for an human being to learn and become a super expert at an language.

The real problem lies in the unwritten culture rules of languages. While there is no easy codification of this, I think given good enough comprehension, storing a few dozen book worth of info should be sufficent, assuming no truely incomprehensible alien species is on the list.

Now consider archologists decoding unknown languages, the amount of information they have is limited to dozens bytes worth of data and some background info.

With only a few hundred species to interact with and common cultural bases across much of the galaxy, this doesn't mean much. Darmok's language was BASED ON FUCKING ENGLISH and the thing couldn't figure it out.

The fact that on the fly translation with basically no previous information is possible implies the amount of new information required to learn a new language is so trivial that collection info by interacting for a few seconds is adquate. The fact it works most of time is unbelievable to begin with. However, this is trek and it probably wouldn't work in more universes as languages would probably be too different for it.

[SWPIGWANG]

Storage would require at least the 100 billion stars to be held. We know that some ships, like X-wings, can skim by close enough to stars that a slower ship would be pulled out of hyperspace by the gravity well of said star (course plotting stuff from X-Wing novels). We also know that slower ships exit hyperspace near planets. Which means, at the very least, that an X-wing can hyperspace right through star systems. Which means each star's satellites must be stored, too. We're getting well beyond 4E10 objects now.

100 billion, thats 1e11 or something right? Now given 100 relevent objects in each star system, that is 1e13 objects to keep track of. Very well.

Now all we need is the required accuarcy to find the amount of data needed to store a table star table.

What's the complexity of a "find the straightest course that doesn't hit anything" algorithm? I haven't found anything applicable, just routing algorithms (would work for multiple jumps, but that's Rebel strategy; those who don't need to hide their location can plot a curved course and use much less fuel) and on-the-fly computations for robotics (which won't work at all, especially with the 90 degree turn things going on).

Well, I guess we can use the pathing algorithm, when assuming space is quesi-static.
http://en.wikipedia.org/wiki/Shortest_path_problem

The number of nodes would depends on the required accuacy of the jumps and just how the spacecraft can maneuver in hyperspace to make up for inaccuacies in initial calculations.

It'd could be along the lines of taking a line and then generating some kind of sine wave function and modifying it for each obsticle it finds to get the smoothest possible course.

That would be assuming that the shortest "physical" path is also the fastest temporally. In anycase, if an algorthm based on this works (depends on hyperspace topology afterall) than calculation is absurdly easy with new objects added on order n (assuming star tables build in).

If one simply wants an okay jump solution, and assuming that spacecrafts can have a dozen or so degree per hour worth of turning (if applicable) in hyperspace(i think it was shown in an EU novel or something for in ISD), it an spacecraft maybe neglect most calculations altogether and simply "dodge" systems as they come up. (space is petty empty after all)

We do know that even stray comets(Dark force rising) and asteroids( Rebel Dawn) can affect hyperspace jumps for smaller vessels. Han comments about supernova and the like represent an alternate end of the spectrum. To put it simply, we have to use super-computers now just to safely send our rocket ships in our own system. Imagine the drastically larger scale needed, to move ships on an inter-stellar basis.

Everything affects everything. An butterfly can affect (or create) an hurricane. However the problem is how much and whether hypersapce jumps take into account of them.

As for supernovas, all you have to do is program them into the droid. There isn't that many supernovas too worry about.

As for super computers for our spacecraft, Apollo era craputers and of course mercury era slide rules were adequate for spaceflight. The thing about getting better computers to do the work is because computers are cheap and better accuracy can't hurt. (as the true optimal path can not be solved exactly, mathmatically, without infinite processing power anyway) In anycase, the computers on say the space shuttle or ISS are all worst than any laptop and astronut can bring with them. (radiation hardening is expensive and spacecrafts really don't need to crunch numbers anyway, just store whatever data crunched from a mainframe is good enough)

-----------------------------------------------------------

That said, I don't think hyperspace directly map to real space in an simple way. I remeber in old X-Wing vs Tie Fighter games (or something) where it is said that each point in space maps to hyperspace that "overlaps" with another point of real space, allowing jumps between two points. While that is probably not canon, but how complex the math would be really depends on how non-linear hyperspace is. If it is something like turblent fluid flow or something that is extreamly nonlinear with complicated time dependence that required processing and storage power goes up to near infinity. On the other hand, if it is simply dodging the stars with an linear 1:1 space-hyperspace mapping, than it is an largely trivial task. Space is empty and simply consider that when you look out into space, you see darkness and not surfaces of another star. (ah, the old paradox of newton's time) If consider hyperspace as an direct mapping, including effects of gravity (gravity lening can hardly be noticed, ever) then simply pointing to the destination is good enough for most cases. (it is not like another jump is impossible or that sublight drives are terriblely slow, especially if you are running away with people shooting at you, after all)

[General Zod]

As for super computers for our spacecraft, Apollo era craputers and of course mercury era slide rules were adequate for spaceflight. The thing about getting better computers to do the work is because computers are cheap and better accuracy can't hurt. (as the true optimal path can not be solved exactly, mathmatically, without infinite processing power anyway) In anycase, the computers on say the space shuttle or ISS are all worst than any laptop and astronut can bring with them. (radiation hardening is expensive and spacecrafts really don't need to crunch numbers anyway, just store whatever data crunched from a mainframe is good enough)

minor nitpick. there's a vast difference between traveling brief distances in space that don't even leave the solar system and traversing interstellar distances. the two simply are not comparable.

[Mad]

100 billion, thats 1e11 or something right? Now given 100 relevent objects in each star system, that is 1e13 objects to keep track of. Very well.

Now all we need is the required accuarcy to find the amount of data needed to store a table star table.

For a minimum value, we could probably assume the values need to be precise to about +/- 1000 km or so.

Well, I guess we can use the pathing algorithm, when assuming space is quesi-static.
http://en.wikipedia.org/wiki/Shortest_path_problem

The number of nodes would depends on the required accuacy of the jumps and just how the spacecraft can maneuver in hyperspace to make up for inaccuacies in initial calculations.

Those aren't what I was thinking of, since that's jumping from node to node. The idea of hyperspace navigation is to avoid as many nodes as possible, since each node is a star system with lots of obstacles to crash into.

And then the cost to each node would have to be stored. Does each star map to every other star? If so, that bumps the storage requirements to a minimum of 1E22 bytes, assuming 1 byte per star. If not, then how many stars away do we link to?

And how do we account for incoming direction? Take this, for instance:

Code: Select all

 D
  \
A-B-C

Sure, you can get from A-C pretty easily, but D can't get to A through B because hyperspace isn't going to allow that kind of course correction. But with those shortest-path algorithms, such cases aren't going to be covered. (Basically, those shortest path algorithms are better suited for Internet packet routing, and not navigation.)

And I'm not even going to get into the computational complexity of them...

That would be assuming that the shortest "physical" path is also the fastest temporally. In anycase, if an algorthm based on this works (depends on hyperspace topology afterall) than calculation is absurdly easy with new objects added on order n (assuming star tables build in).

Actually, what I had in mind was "what would require the least amount of fuel." If you wanted to take the fastest route, and if hyperspace doesn't map 1:1 with realspace, then even more number crunching would be required.

If one simply wants an okay jump solution, and assuming that spacecrafts can have a dozen or so degree per hour worth of turning (if applicable) in hyperspace(i think it was shown in an EU novel or something for in ISD), it an spacecraft maybe neglect most calculations altogether and simply "dodge" systems as they come up. (space is petty empty after all)

Except at the speeds we're talking about, trying to dodge one as it comes up will require you to start turning so far away that you'll risk hitting a different star. It's way, way too risky to do it that way.

Everything affects everything. An butterfly can affect (or create) an hurricane. However the problem is how much and whether hypersapce jumps take into account of them.

A butterfly can affect a hurricane, but the effect is negligible. I doubt a butterfly's mass should be taken into account when doing calculations on landing on Mars, for instance, even though it generates enough gravity to have a very small non-zero force on everything in the universe. Thus, it can safely be rounded to 0 with no noticable effect.

[SWPIGWANG]

Those aren't what I was thinking of, since that's jumping from node to node.

The idea is to map space into a pathing tree. So for example we can have say, 1 light year x 1 light year blocks of nodes and calculate the time/fuel needed to get though. I was thinking that the nodes don't have to be stored but simply calculated on the fly and that most nodes can be clipped by heuristics.

Direction considerations can be done when building a new tree to take into account for it. Impossible path would simply be assigned infinite cost different costs can be assigned for different directions. The cost between the nodes can be used to optimize fuel, speed and things like possible interdiction and such.

Except at the speeds we're talking about, trying to dodge one as it comes up will require you to start turning so far away that you'll risk hitting a different star. It's way, way too risky to do it that way.

Wouldn't the worst thing simply to drop back into real space, reorient and than jump again? In anycase, dodging is not too difficult because space is space and the number of physical objects is almost nil. The mean free path is effectively infinity after all. You are not going to hit a start anytime in your lifetime even if you travelled all your life, statistically.

[Mad]

The idea is to map space into a pathing tree. So for example we can have say, 1 light year x 1 light year blocks of nodes and calculate the time/fuel needed to get though. I was thinking that the nodes don't have to be stored but simply calculated on the fly and that most nodes can be clipped by heuristics.

Direction considerations can be done when building a new tree to take into account for it. Impossible path would simply be assigned infinite cost different costs can be assigned for different directions. The cost between the nodes can be used to optimize fuel, speed and things like possible interdiction and such.

But the path through each 1x1x1 block will be different depending on the entry vector and location. How many entry points and vectors will there be on the blocks near the target point, for instance?

Wouldn't the worst thing simply to drop back into real space, reorient and than jump again?

Horribly fuel ineffecient. Each jump requires a large amount of fuel, while staying in hyperspace only sips it (Ref: X-Wing: Rogue Squadron).

In anycase, dodging is not too difficult because space is space and the number of physical objects is almost nil. The mean free path is effectively infinity after all. You are not going to hit a start anytime in your lifetime even if you travelled all your life, statistically.

Gravity wells, on the other hand, are different. Can't travel too close to one or you'll get pulled out. (Basically, I'm envisioning what happened in Larry Niven's "At the Core" using a much slower hyperdrive than what is used in SW. Constant dodging, while at the rim.)

[SWPIGWANG]

But the path through each 1x1x1 block will be different depending on the entry vector and location. How many entry points and vectors will there be on the blocks near the target point, for instance?

uh...I guess that wouldn't work than, since the algorithm complexity would go to hell....

ah, it was a attempt anyway.