News for hackers. Stuff that matters.
Re:The Blink Protocol by cnvogel on Sunday January 20, 2013 @08:45AM (#5086861)
Also in applications where bit-errors (or disconnections) are to be expected, the <length><data_of_length_length> ... approach makes it hard to re-synchronize after errors. (think of a serial cable that might be dis/reconnected, or radio-frequency links.)

A more robust encoding reserves some bit-pattern as a <start-of-message> marker, and escapes regular appearance of this marker in the payload with some special sequence. (see, for example, http://en.wikipedia.org/wiki/High-Level_Data_Link_Control ).

When you start putting (for example) single "blink"-messages in distinct payloads in such a framing container, then the <length-of-message> byte of blink becomes redundant, of course.

Re:The Blink Protocol by randersson on Monday January 21, 2013 @12:40AM (#5090067)
We discussed length preambles vs markers when we designed the FAST protocol 8 years ago. That time we made length preambles optional (there are two "modes"). The nature of the FAST encoding made it impractical to use markers even though it would have been entirely possible. This time around we decided to use length preambles. Blink focuses on encoding short messages, so streaming works well with preamble. We expect bit-level errors, disconnects etc. to be handled by a lower layer. Rolf