RE: How to limit file revisions

[Arthur Barrett]

Jim,

>CVS breaks all files, whether binary or text, into "lines" 
>where a line is roughly defined as "any sequence of bytes 
>preceded by ASCII LF (0x0A)". Thus, all files store only 
>the deltas required to generate the previous revision. 
>In the worst case, a file may contain no LF characters 
>and thus the entire file must be stored. Does CVSNT behave
>differently?

My explanation was a little simple - but that's about as much as I know.  CVSNT 
uses an alternative algorithm when -kB is used.  The efficiency of the text 
based algorithm on binary files is relatively poor (especially when compared 
with its efficiency at storing text files).

More info is here:
http://www.cvsnt.org/manual/html/Binary-howto.html

CVSNT also can store Unicode files as text and the text diff algorithm 
correctly handles UTF-16 etc (which is common in Windows).   

I suspect the original poster may be storing Unicode files as binary (which is 
required to not corrupt them in traditional CVS), and they would be more 
efficient as -ku files in CVSNT (and then the users could diff/merge as well).

Or they may just be frequently modified binary files...

Regards,


Arthur