http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5167
------- Additional Comments From [EMAIL PROTECTED] 2006-11-07 02:05 ------- I understand what you're saying, but doesn't this mean that Bayes is making assumptions about the storage backend beyond the interface specification ? It's effectively doing: - give me a read-only instance of the storage; - do an update (assumption: it's ok because I know the backend will defer this until later when I call sync in a read/write context...) This isn't a safe assumption (although it is true for DBM backends) unless the interface specification is changed so that touch_tok_all becomes semantically schedule_touch_tok_all - for example the SQL backend doesn't use any kind of journal but it gets round this by not (really) differentiating between ro and rw connections. I could implement a journal but at the moment I'm thinking about making tie_db_*() a no-op and handling connections internally. That might actually be a neater approach for what I want anyway. I worry that I might be missing something in the greater OO architecture of SA though since I'm fairly new to it. ------- You are receiving this mail because: ------- You are the assignee for the bug, or are watching the assignee.
