[freehaven-dev] Actually, no, we have information dispersal.

[dmolnar@belegost.mit.edu]

My mistake. Looks like the code is doing IDA after all, but just didn't
notice on such a small file. 

This time I tested it on one of the object code files created during make.
Same scheme - 10 shares, any 5 reconstruct. Notice that each share is
much smaller than the original file. 

-rw-r-xr-x   1 dmolnar  freehave    28012 Mar 17 16:11 cryptlib.o
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.00
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.01
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.02
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.03
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.04
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.05
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.06
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.07
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.08
-rw-r--r--   1 dmolnar  freehave     5640 Mar 17 16:35 cryptlib.o.09

I tested with shares 00-04 and reconstructed the file
/usr/home/dmolnar/crypto30/aregood . diff says that this and cryptlib.o
are the same. 

Now to be anal, we should probably try all subsets of 5 servers which
could be lost...but I'll worry about that _later_. 

Thanks, 
-David