Bug in calculating ISIZE for REORGCHK on indexes??

[Amit]

Hi:
I'm trying to manually compute the ISIZE in REORGCHK for indexes in
DB2 v8 (fp 4).

Based on the documentation in Administrative Guide for v8, for non lob
fields
ISIZE should be the aggregate of avg datasize of the index columns + 2
bytes (1 byte for v7) overhead for each varchar and vargraphic column
and 1 byte overhead for each null column.
I use AVGCOLLEN in syscat.columns for the column size, and add 2 bytes
for each varchar/vargraphic column and 1 byte for each column that
allows nulls. The result is a lot bigger than the ISIZE in reorgchk in
DB2 v8.

If however I subtract 2 bytes from AVGCOLLEN for each
varchar/vargraphic column and ignore the adjustment NULLS attribute
for the column, my result matches REORGCHK.

I also found IBM APAR Software JR18802: REORGCHK OVERESTIMATES THE
ISIZE (fixed in fp 3) that seems to suggest that REORGCHK was adding
extra bytes to AVGCOLLEN which is already corrected for VARCHAR/Nulls.
If REORGCHK were indeed adding the extra bytes before, and this
problem was corrected, why does my subtraction of 2 bytes from
AVGCOLLEN for each varchar column give me a result that matches
REORGCHK? This APAR seemed to imply that AVGCOLLEN accounts for all
the overhead with varchars and nulls and ISIZE can be just computed by
summing the AVGCOLLEN's of all the columns.

Any clarification on whether I need to adjust the values in AVGCOLLEN
in order to compute the correct ISIZE will be appreciated.

Thanks
Amit