From: Alex B. <boi...@in...> - 2006-03-30 20:07:38
|
Ok, just wanted some clarification. If you believe you'll have a need for them, let's have 'em. I'm also behind a more capable architecture that you describe. alex Thompson, Bryan B. wrote: >No, I have not identified a need for intention/implicit locking for btree or >htree. However I am looking to support (and use) more than a flat resource >model myself, so that is a strong motivation for me. > >Jdbm has a history of small, light and tight. Keeping to that profile is >important for many uses, but I am interested in a more extensible >architecture since I want to do more with it. I think that a framework >layer supporting collections, one-to-one, one-to-many, and many-to-many >associations, indices over attributes in collections, etc. would be a >typical consumer of an extended locking model such as I am proposing. That >is certainly what I want to do with it. > >I assume that you want to do deadlock detection, even with just a flat >resource model, right? What do you see as the minimum locking model then, >an integration of a multiple-readers, exclusive writer lock on a per >resource (record) basis with an online deadlock detection algorithm? > >-bryan > > |