[tex-live] volunteer to implement lcab compression?

[Olaf Weber]

Manfred Lotz writes:
> On Tue, 8 Jun 2004 19:33:55 +0200
> Thomas Esser <te at dbs.uni-hannover.de> wrote:

>>> We (Sebastian and I) have not found any other programs to write .cab
>>> files under Unix (many programs support reading .cab's).  There is a

>> Is the file format good enough to preserve file attributes such as
>> UNIX permissions? 

> Cab format is intended for Windows platform. Therefore it doesn't know
> about permissions and ownership. 

The format allows for additional data to be stored per cab file header
(CFHEADER), folder header (CFFOLDER), and data block (CFDATA), but not
per file header (CFFILE).

However, the additional data in a folder header is the same size for
all folders in the CAB file.  And the same goes for the additional
data per data block.

A data block can contain data of more than one file.

All of which taken together means that there is no really suitable
place to put additional per-file data in a CAB file -- you'd have to
resort to indirect methods like packaging an additional "file" with
the data.

In all it is not a very good distribution format for unix-like
systems.

>> Is the file format smart enough to handle text files
>> properly(i.e. convert line ends on extraction)?

> I'd say yes because CABARC.EXE seems to treat files as binary. 

Um, this means "no" as Thomas was asking about automatic translation
of NL to CRNL or CR, and all that.  And the answer is that for CAB
extractors the CAB file really contains a number of bytestreams, which
are not interpreted or changed.  (Personally, I like that.)

Basically, CAB does some tricks to get better compression than ZIP
will for the case with many small files, and is as far as I can tell
about equal on large files.

-- 
Olaf Weber

               (This space left blank for technical reasons.)