4 ms·
Could you tell me whether he's suggesting to send a batch of data packets, then wait a 44-minute turn-around time to hear back from Earth which ones failed so t
by jessriedel 9y ago
Could you tell me whether he's suggesting to send a batch of data packets, then wait a 44-minute turn-around time to hear back from Earth which ones failed so the failed ones can be resent? This is almost certainly not the best way to do this, from an info-theory perspective, right? Wouldn't you put the entire batch of data in an error-correctable code that you could ensure had very high likelihood of being fully recovered on Earth? Send-check-failure-resend works when you have a fast but intermittently congested network on Earth, but it seem silly for a connection with such extreme latency.
- greglindahl 9y ago100% of current data from deep space uses FEC, so you're asserting that his brief answer says that we should stop doing the usual stuff? Tough crowd.
- jessriedel 9y agoMy question is an honest attempt to understand, not a veiled criticism. I don't know what FEC or the usual stuff is.
- greglindahl 9y agoForward Error Correction involves sending extra data in the initial transmission so that most errors can be corrected without needing a resend. It sounds like that's what you were attempting to describe in your comment. https://en.wikipedia.org/wiki/Forward_error_correction https://en.wikipedia.org/wiki/Forward_error_correction FEC is usually used in addition to resending.
- jessriedel 9y agoThanks! If FED is standard for deep space and Musk know this (both reasonable), do you have any idea what he might have meant by "then do a batch resend of the CRC-failed packets"? Is it just that guaranteeing zero packet losses has a large overhead (surprising to me), so that the optimal FED protocol in deep space does in fact include a bit of resending?