>>48
It's not inflexible because if something can't be encoded in a dictionary there is just a fall back to a more typical variable length encoding based on character set, it's additive. Whether or not it would be costly to process seems to depend on the application. Any sort of manipulations of words such as spell checking and word based search would be faster, as would transmission over a network. The size of the lookup table would not matter for anything but embedded machines, which shouldn't be doing text processing anyway. Character based manipulation would be slower, but in most cases it probably shouldn't be used anyway. If you have efficient structured binary encodings of source code, logging, and protocols what would you need efficient character based manipulation for?