The literature has always been very vague on this subject. Regarding datasourcs like 1_CO_PA_OC_1000, there are 2 options for initialising the deltas:
Init WITH data transfer = FULL LOAD and Set delta pointer (at end of full load)
While the full load is taking place, there is a risk of missing changes before delta record is set, resulting in missing data
OPTION 2. If I do the following sequence:
1) Init WITHOUT data transfer
2) Full load
3) Then run delta packages after
The delta pointer is set BEFORE the full load. So long as the full load is imported to BW before deltas, then there should be no data loss, because the subsequent deltas should load incorporate changes during the full load.
I am loading into DSOs and later cubes.
Is there no risk of data loss with OPTION 2, so long as full load done before other deltas, or is there some disadvantage with OPTION 2? When is OPTION 1 preferable?