For specific types of data or usages the loss of data must be avoided in any case. Examples for that are executable program files, text files or original multimedia files intended for archiving. In the same way unknown data types have to be handled because it is impossible to define acceptable variations and losses in quality.

Compression procedures belonging to this catagorie try to use the best possible form of data representation. Therefore a large variety of opportunaties is available:

Frequent symbols will be encoded by short codes, rare symbols by long codes.
Sequences that were already part of the former contents will be replaced by addresses.
Dictionaries will be created that contain previously encoded sequences. The algorithm addresses the dictionary instead of the contents.
Repetitions will be indicated by counters.
etc.

< ^ >