Cart
[Home] [By Thread] [By Date] [Recent Entries]
A decently designed binary data encoding protocol will take less space to encode and transmit, and less time to decode, than a decently designed text based one. Just think: writing an integer as four bytes takes one instruction, maybe two if byteswapping is needed. But writing "32768" takes more bytes, and more time to encode/decode. Text compression costs even more. There is no way to recover those differences. Of course those "decently designed" binary protocols rely on things like contextualized encodings. If you demand self-descriptive data, then it's easy to get to the point where manipulating type data costs so much that encoding/decoding differences are in the noise. And some programming environments basically don't support binary data worth speaking of. So there are going to be tradeoffs to consider. - Dave
|
