2 replies to this topic

[treonsverdery]

anybody here think that this goes well with multicore processors to double (or higher) internet speed

First lets compress some data.
Think of wikipedia as a list of digits possibly starting with 1649121 (then on).

a computer program that sought to batch together groups of digits, like"16", "121", "49" then find their square root could create a lossless compressed version of wikipedia noting that each of the numbers has a whole root 49 -> 7*7 , 16 ->4*4 121->11 , Thus the compressed version would start with 7 4 11

now of course the computer program would batch together noticeably longer groups of digits to find a whole number square root, with either an algorith or possibly a lookup table noting things like 393129 -> 627 *627 or 139679345169 -> 373737*373737 this is particularly applicable to multicore processors as one of the multiprocessors can do the expansion of the compressed file.

Now going on with compressing wikipedia so it can travel the internet losslessly with half the packets. after the first pass on the wikipedia file, finding all the easy roots, then the (it would be an easy root if we added 1), or (it would be an easy root if we added 2) or possibly (it would be an easy root if we used a cube root, or a 5th root) versions would be described. Also algorithmically you can see pretty easily what to add, like so
61134.06363885849 squared is 37373737, algorithmically to get rid of that pesky .06 you could use a loop to narrow towards a whole digit, subtracted from the source file, which is then kept at a separate array list of modifiers.

It seems that someone that knows math might say that "the thing after the decimal can be removed if you add negative .06...*3737373737 to the source compression number, then note a (modification feature) at an array" I think these are a bunch of multiprocessor friendly possible computations that map out the entire wikipedia digit sequence as batches.

The benefit is that this compression increases the speed of the internet at what the lossless compressed files compression ratio is, this is also packet compatible as it is brief areas of batched digits that have been compressed. also the source file only needs to be compressed once, thus can be chunked at multiprocessor friendly batches like pages.

I thought of this as a way to support the singularity as doubling the speed of the internet supports the singularity.

Edited by treonsverdery, 16 November 2012 - 11:58 PM.

[8bitmore]

Nice thinking; someone has long ago implemented this in a very efficient manner in the form of server-side gzip compression which is already being used almost universally. Wikipedia pages are all already gzip compressed and yielding an average improvement of about 70% compression per page. You can check for yourself using:
http://www.whatsmyip...mpression-test/

For a little bit of meta fun I tested the http://en.wikipedia....TTP_compression URL and got the following results:
Original Size: 43.26 KB
Compressed Size: 12.17 KB
Data Savings: 71.87%

[AOLministrator]

I am a sysadmin and as such I can confirm that there is only one feasible way of doubling your internet speed, meticulously demonstrated in this video:




Regardless, just read about compression. In simple terms, you need to feed it sufficient data, that induces delays and such, very ugly. Many 3G providers support deflate compression or even default to it (afaik), and if you have ever used it you will know it sucks badly and only has disadvantages (except reducing the providers bandwidth expenses).