Hi Patrick, nice idea, and thanks for suggesting it. I have not had the opportunity to look closely at Schevo, i.e. use it, and I became aware of it only when it was pointed out here around pycon time. I was already planning to package a first release of moellus at that time. Schevo is certainly interesting, and definitely more "ambitious"... implementing rigourously the idea of an object model, plus accomodating several back-ends, multi-editing tools, and so on. As you say, if I were to take moellus that route, it would probably be a better idea to take on Schevo instead. For me moellus was simply a means to simplify usage of durus. And I have no intention of taking it much further. I mention port to zodb below only in the spirit of "genericity", asking what if? I really never had the intention to add zodb support to moellus, and still don't... for using zodb there is a lot more knowledge out there, than for using durus, and I would have looked at how people use zodb first. The driver behind moellus is the application I am working on... and the features have evolved to meet those needs. A basic stumbler would therefore be to port the application to another db layer. For new project opportunities, Schevo is certainly on the my list of tools to be considered. I am not even familiar enough with Schevo to be able to compare features... there is obviously some overlap, but i suspect they are very different animals, and probably there are problems that may be better addressed by one or the other. Just as there is good justification for a db like Durus, when ZODB exists... or for Quixote for that matter. If there is one problem with Python, it is this, that it is so easy to implement things yourself! Anyhow, I am very attracted by small light-weight designs, and my intention is to keep moellus as light as possible. But, really, I think it is silly, if flattering, to compare these two frameworks... Thanks for the links... weekend reading ;) Cheers, mario On Jul 8, 2005, at 2:50 PM, Patrick K. O'Brien wrote: > Mario Ruggier wrote: >> >> I am using persistent instances **liberally** in btree keys, so a >> straight porting to zodb would not work. However, the way my >> persistent >> items are defined, i can very easily get a (humanly descriptive) >> string >> tuple to identify each item.... but, that will add the overhead of >> translating of instances to such immutable objects, on all the gets... >> well, i guess it can always be another computed attribute ;) > > Perhaps you'd like to join forces in working on Schevo. We've been > doing a complete rewrite of Schevo from the ground up, specifically for > Durus. I believe we've already solved many of the issues you will be > facing in trying to bring a relational model to Durus, since we've been > working this issue for almost five years now. In any case, our latest > code can be found here: > > http://svn.orbtech.com/cgi-bin/schevo.cgi/browser/branches/rebel/ > > Our Durus-specific code is here: > > http://svn.orbtech.com/cgi-bin/schevo.cgi/file/branches/rebel/schevo/ > database/duru.py > > -- > Patrick K. O'Brien > Orbtech http://www.orbtech.com > Schevo http://www.schevo.org > Pypersyst http://www.pypersyst.org >