AZip 2.25

AZip 2.25 screenshot - click to enlarge

  • Added buttons for New, Open and Toggle view commands.
  • Cases where Zip format capacity is exhausted are properly handled.
  • Sorting is a bit faster.
AZip Web site: http://azip.sf.net


TeXCAD 4.51

Just removing some dust on this project...

The latest change is purely cosmetic but useful: buttons are better visible in today's very-hi-res screens.

TeXCAD screenshot - click to enlarge

TeXCAD is a simple LaTeX {picture} drawing tool. Free, open-source.
Web site: http://texcad.sf.net

Zip-Ada v.55

Changes in v.55:
  • Zip_Streams: ZS_Size_Type is now 64-bit signed, enabling Zip.Create to capture archive size overflows in Zip_32 mode.
  • Zip.Create raises Zip_Capacity_Exceeded when archive creation exceeds the Zip_32 format's capacity: 4GB total size, 65,535 entries.
  • Zip.Create is now using an Ada 2005+'s Containers's Hashed Maps; creation is much faster on Zip archives with many entries.
  • (Tools) ReZip has a new option for working only with its own internal compression algorithms - those provided by Zip.Compress. This option is useful if external tools are not available.
  • New Trained_Compression package: generic streaming encoder-decoder engine with the capability of training the engine with data known in advance, in order to achieve better compression. Not Zip-related.
  • Minimum required Ada version is now Ada 2005 (was Ada 95 before).



One-year USD LIBOR rate at 3.13% !

Almost three weeks (!) after us, someone at ZeroHedge has noticed the spiking LIBOR rates in US Dollars. Link to ZH article here.

The one-year USD LIBOR rate is now at 3.13%, not seen since the tumultuous 2008 autumn!...
At the time, the debt card castle was much smaller, and it was crumbling.
So... Well... What to say?... Perhaps: "Good luck" ?

Of course ZeroHedge has better information about the implications, and we learn in their article that there are $200 trillion of debt in USD, directly or indirectly exposed to the LIBOR rate.
This is a little bit more than the figure from here, citing the Fed, about directly LIBOR-linked debt: $7.5 trillion.


Trained compression

For improving compression, it is possible to train a compression scheme with data which is known in advance to both the encoder and the decoder. If the new data to be sent is similar to the training data, you can get in some cases impressive compression gains.

You can see this effect "manually" by doing:
copy /b data+trainer appended
(or the equivalent UNIX command), then:
zipada -el3 trained.zip appended
on one hand, and:
zipada -el3 untrained.zip trainer data
on the other hand.
You'll see that "trained.zip" is smaller than "untrained.zip", although the contents are the same. If you need to send many "data" you don't yet know exactly, but that you know they will be similar to "trainer", you can leverage this information by informing once for all the encoder and the decoder with the "trainer".
While this technique is available with some compression libraries, we provide a generic, standalone Trained_Compression package on which you can plug any of your preferred compression schemes. An implementation is provided with our (equally generic & standalone) LZMA[.*] packages.
Check the "trained" directory here or here (revision 700+).

Here a schema of how it works:

Some results are shown here:

Bullshit Generator: the Next Steps

Here we admire again the recyclability of high-quality bullshit. The latest additions to the  CBSG  stem from an article written in 1996 - a...