I wouldn't include the escape chars added to escape the CRC itself.
I think it's inconsistent to leave the P and D bytes, which are "framers",
but not include the escape chars -- if they are part of the frame and not
the packet.
Also, I think the difference between taking a CRC on the frame+escape
chars and not, is that if one was going to try to do an error correction
using the CRC (I think 1 bit can be corrected), you could NOT do it if you
didn't include the escape chars. The problem is that you wouldn't know if
the error was mistakenly using escape when none was there, or if the error
was in the packet itself. If you included the escape chars, you remove that
extra layer of confusion.
Regards,
Kurt
----Original Message Follows----
From: Jeongyeup Paek <[EMAIL PROTECTED]>
To: KURT PETERS <[EMAIL PROTECTED]>
CC: [email protected]
Subject: Re: [Tinyos-help] Making heads or tails of TEPs113, 111, and actual
code
Date: Tue, 13 Nov 2007 00:12:36 -0800
Well, imho, I don't think there is any difference in error-detecting
capability whether or not the escape chars are included in CRC...
but...
what if the 'calculated crc value' needs char-stuffing?
Should crc add escape char generated by itself?
Wouldn't it be cleaner if not?
Thanks
- jpaek
KURT PETERS wrote:
Thanks for all your help. Here's a semantics question. It seems that CRC
is "frame level" as you said, but when I look at the code for serialP.nc,
it looks like the CRC is only done on the packet level. That is... the CRC
is calculated, then the byte is sent to the "frame layer" where an extra
escape character (0x7d) can be placed into the stream (see code snippet
below). SerialFrameComm.putData(txBuf[txIndex].buf) seems to call the
HdlcTranslateC.nc putData which adds the escaped sequence, when necessary.
If CRC is truly a frame CRC, the CRC calculation should include the
escape byte for escaped data, in my humble opinion.
case TXSTATE_INFO:
atomic {
txResult = call SerialFrameComm.putData(txBuf[txIndex].buf);
txCRC = crcByte(txCRC,txBuf[txIndex].buf);
++txByteCnt;
if (txIndex == TX_DATA_INDEX){
uint8_t nextByte;
nextByte = signal SendBytePacket.nextByte();
if (txBuf[txIndex].state == BUFFER_COMPLETE || txByteCnt >=
SERIAL_MTU){
txState = TXSTATE_FCS1;
}
else { /* never called on ack b/c ack is BUFFER_COMPLETE
initially */
txBuf[txIndex].buf = nextByte;
}
}
else { // TX_ACK_INDEX
txState = TXSTATE_FCS1;
}
}
break;
Your thoughts?
Regards,
Kurt
--
Jeongyeup Paek
Ph.D. student
Embedded Networks Laboratory
Department of Computer Science
University of Southern California
http://enl.usc.edu/~jpaek
_______________________________________________
Tinyos-help mailing list
[email protected]
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
