 |
Branche_Ungrouped.png (108,695 bytes) |
|
|
BankEmployee.png (13,258 bytes) |
|
|
Where the reducing of Branche_Ungrouped may seem the start of the issue, it really starts in the elementary model at BankEmployee. CaseTalk already warns for: - OTFT's Branche, City, County, State may not be verbalizable - Binary facts StateName, CountyName, CityName may not group two ways These warnings should prevent the modeler to GLR, since it'll lead to apparent subset loss. First of all: the binary facts mentioned will not be reduced in the newer CaseTalk. It'll remain ungrouped. Which leads us to State,County,City which are candidates for Reduction. The following subsets will remain: SC 10->7 and 5->7. Yet without the presence of a totality constraint on State, the State will disappear completely without a trace of subset constraints. This will cascade over County, State and County. Even towards Employment, which will end up with a single subset towards Bank. This combination of effects is the true cause for foreign key failure. How could CaseTalk warn for this, or prevent it? Should it stop Reducing State, County and City if they are not verbalizable? Or is there something else we can think of? |
|
|
Does this still apply after the report and future fixing of bug 0002647. |
- Anonymous
