1

Topic: Dispersion of length of archive?

If to take some dial-up independent (from each other) bit sequences of length N, to pack all of them what  the archiver (for example, zip) and to count mean value (S) lengths of the turned out archives and a dispersion (D) lengths of archives...
1. What (in the order of value at least) will be relation D/S?
2. At what of known archivers (along with zip) this relation will be noticeable"big (than at remaining)?
3. By the way, whether there will be a allocation of lengths of such archives (for the specific archiver) normal... That is, type, what there - Gaus or the Poisson?

2

Re: Dispersion of length of archive?

Ivan FXS;
1. It is necessary to take not a dispersion, and a root from it. Its relation to an average will be called as coefficient of a variation and dimensionless.
2. The experimental check occupies minutes 10.
3. Most likely, anything similar to the "correct" allocation will not be.

3

Re: Dispersion of length of archive?

FXS wrote:

1. What (in the order of value at least) will be relation D/S?

It is a senseless index, at numerator and a denominator different dimensionalities.
And generally - a question in itself too . Archiving as process leans against that fact that the real data possesses an inner pattern and are essentially not casual. Attempt to check up something on any casual dial-up of bits leads only to the answer to a question "what archiver adds less  to the incompressible data".

4

Re: Dispersion of length of archive?

About! Never checked,  plaid about  compression, it seems a theoretical limit of compression ~2/3 (for texts ~ 1/4). It is possible on an ehe-file example. Redundancy .
I Assume that and will be.
I Will add Sokolinsky. Yes, a typical measure = M/sko, instead of on . .,   finds room in M.
Such wide market of archivers? In Puassona/Releja somehow on sense it is more trusted: it is not enough strong algorithms, feeble - it is more, and nobody uses bad. Type so it is expected. Truth it for pure compression (not for a file)
It is necessary to remember that in oblate archives certain kol-in goes on , and even then the empty archive can appear not zero.
And generally, it is all was o-o-very much for a long time...

5

Re: Dispersion of length of archive?

And if to corrigendas not to cling (it is finite, the root from D, instead of D), and under the content something to tell to try? Not?
______________
softwarer, I unless told "a casual dial-up"?
________________
And yes, the question could be formulated so:
1. How statistics of compression ratio (for "typical" archivers) looks?
2. At what archivers of statistican of compression ratio has big (in comparison with other "typical" archivers) a dispersion?

6

Re: Dispersion of length of archive?

Here something was.
http://www.compression.ru/arctest/act/act-calg.htm

7

Re: Dispersion of length of archive?

exp98 wrote:

About! Never checked,  plaid about  compression, it seems a theoretical limit of compression ~2/3

It seems to me of a theoretical limit of compression is not present. For example  a file from one zero it is possible to compress "terabyte of zero" to a word combination.

8

Re: Dispersion of length of archive?

WebSharper, well it too from area of cavils. Then and texts it is possible to compress about one link to a global corpus. Speak about a statistical theoretical limit generally.  to take - it will not be compressed, therefore as there already Hafman rummaged. And take  a placenta-st... And still I remember that English is compressed more than Russian, since there language for "persons with restricted abilities". For certain is , specialized under a reality.
And I interpreted a question of the HARDWARE as for casual .

9

Re: Dispersion of length of archive?

exp98 wrote:

WebSharper, well it too from area of cavils. Then and texts it is possible to compress about one link to a global corpus.

So.
[quote =] Speak about a statistical theoretical limit generally.

You could not make determination of it?
[quote =] since there language for "persons with restricted abilities".

Zadornov is live!
[quote =] And a question of the HARDWARE I interpreted as for casual .

Their casual samplings of that?