30 Jan
2009
30 Jan
'09
7:42 p.m.
Tres Seaver wrote:
Is there any reason to hide the KeyError behind a 500, rather than letting it propagate to the application? In this implementation, we lose the information about the bad 'status' value.
I agree, but I hit a legacy design snag. The handleException() method still calls setStatus(exception_type_name), where we could be handling any kind of exception, similar to what Zope has always done. IMHO, we should change the contract of setStatus() to not do any lookup at all and make handleException() do the lookup instead, or something along those lines. But I'm trying to make conservative changes ATM. Shane