[squeak-dev] Reed Solomon Blues

Robert Withers robert.w.withers at gmail.com
Mon Dec 7 12:20:15 UTC 2015


Hi Good morning,

I am trying to port java code implementing Reed-Solomon encoding with a 
GaloisField and polynomial system. I am currently testing the easiest 
Mode, which is 9 data symbols, 15 code symbols and 3 symbols of error.  
Theer's 2 nibbles to a byte, so a split and join is needed. This all has 
to do with iteration where the base indexing changes from 0 to 1. I 
screwed it up in there somewhere and I am struggling to find teh issue. 
It looks like a complete code review...

Since this is core crypto code, it would be an extra blessing if someone 
else's eyes were on this code in detail.More tests always help too. 
Would someone be willing to crawl into the java and squeak code to 
review, learn and qualify/validate? Also, you would be helping me 
understand all this polynomial math and where my indexing issue may be.

The failure in Squeak is in the GenericGFPoly>>#divide: method. It runs 
forever and the quotient/remainder never reduces it's degree. So one of 
teh math operations is broken & failing. You can run the EncodingTests

I am including a zip of the Java classes I use. This core Reed-Solomon 
code was written at Google and released open-source. I wrote a test and 
they pass, so code is working. The latest Cryptography package in the 
crypto repo is version 48. Please load that and run the 
CryptoReedSolomonTest to see the infiniteLoop of the polynomial divide.

Let me know if you can help.

Thanks,
Robert

-------------- next part --------------
A non-text attachment was scrubbed...
Name: reedsolomon-java.zip
Type: application/zip
Size: 24035 bytes
Desc: not available
Url : http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20151207/4fd8f40e/reedsolomon-java.zip


More information about the Squeak-dev mailing list