From: Tatu S. <cow...@ya...> - 2008-02-03 17:26:38
|
--- Sam Halliday <sam...@gm...> wrote: > Hi all, > > I was just wondering if this project is still > active? It looks great! Hi Sam! I think it's fair to say that project is not currently active. As far as I know, it is just due to lack of time (as is common with open source projects), although I think there is lots of interest for the library itself. > We have used BerkeleyDB internally for some projects > but got a scare > when we realised it was GPL (luckily we'd never > distributed anything). ... Hmmh. Really? I thought BDB-JE at least did have something more liberal available... it certainly is used in-house at lots of companies that would never touch anything GPL. Anyway, one comment: ... > If anybody is still subscribed to this list... I'd > like to know a > little bit more about the serialisation procedure. > The Serializer > class is somewhat similar to Binding in BerkeleyDB, > but BerkeleyDB is > very smart when it comes to using Java's > serialization. You may have > noticed that if you inspect a serialized object, > there is a *lot* of > data at the beginning which Java uses to store class > name info etc. In > BerkeleyDB, all this "boilerplate" is stored in a > separate table and > that means only the important pieces of data are > serialized to disc, > with no repetition. Has JDBM looked into this? I think it is important that there is clean layering between low-level storing of binary key/value pairs (with possible partitioning by different 'tables'), and more advanced object-binding/serialization functionality. Latter can be built in terms of former, possibly even so that it can be plugged on different backends. So it'd be neat to have a sub-project (or related project) that implemented this functionality. Currently there are classes within JDBM that implement serialization for specific containers, but with slightly different goals than generic serialization. One other thing: I don't know if you have looked at it, but Carbonado (http://carbonado.sourceforge.net/) is one interesting solution to such serialization problems. And has more advanced versioning/upgrade facilities than what BDBs object store has (there is a comparison chart somewhere that shows differences that matter). This comes handy when projects evolve, and new data needs to be added... which is the case for about any succesful project (failed ones tend to remain more stable :) ). While it doesn't have JDBM backend (I think), it does have ones for BDB and BDB-JE, as well as relational DBs. And it might not be very hard to extend it to allow use of JDBM. Anyway, just a thought for something else to perhaps look at, -+ Tatu +- ____________________________________________________________________________________ Never miss a thing. Make Yahoo your home page. http://www.yahoo.com/r/hs |