Mon Sep 14, 2020 11:30 am
It's really up to application. The main difference between "normal" UDP and UDP-lite is that damaged packets don't get dropped with UDP-lite. That application can do with damaged packets is up to application itself.
The idea is that bit stream, as produced by application, has some redundancy built in, e.g. due to use of FEC (forward error correction). Which means that some bit errors can be tolerated because built-in redundancy is enough that receiving app can reconstruct original multimedia stream. Another idea is that there are multiple pseudo-streams interleaved and only the most significant stream really needs to be decoded flawlessly to give useful output, therefore that pseudo-stream will contain significant amount of FEC data. The rest of pseudo-streams contain less significant data (in audio streams it might include stereo information, or frequencies higher than 15kHz, ...) and if those are not decoded properly, rendered stream (either audio or video) is still usable to humans. Thus these pseudo-streams can be created with less (or none) FEC data.
The described implementation doesn't need any payload protection (by means of CRC), only protocol overhead has to be CRC protected (e.g. not to deliver contents to wrong receiver). And UDP lite allows to CRC-protect only a portion of whole IP packet.