Without dictionary, AUL cannot recover rows to dmp format. But we could manually create them. We will talk about segment information stored in "AULTAB.TXT" file, the file name must be in upper case under Linux/Unix. The text file should contains the following columns (split by ",") :
1, Object ID
2, Data Object ID
3, Tablespace ID
4, The relative file number of segment header
5, The block number of segment header
6, The create order in cluster if table is clustered
If you have system tablespace left, it could be generated by "unload table tab$;" command. This command will read rows from TAB$, IND$, TABPART$ etc, to get the data object id and segment header information, please run this command after you get the object information by "unload table obj$;" command. For example:
12,6,0,1,57,1
32,29,0,1,177,2
22,10,0,1,89,1
18,18,0,1,121,
4,2,0,1,25,1
AUL will get object's data object id and segment header address from this file, and the locate the segment header and get the extent allocation information (Extent Map) to perform recovery. When there is "AULEXT.TXT" exists on the same directory, it will just get data object id from this file, and get extent allocation information from "AULEXT.TXT" file, so if you have a wrong "AULEXT.TXT" file, it could affect the recovery result.