I'm glad this thread has started some discussion; it's good to get some "soundings".
To be clear: the proposal isn't about removing NO or autogenerated UIs, though it is about making sure that code that is incomplete is clearly marked as such so that would-be users don't try something out, find it doesn't work properly, and then walk away. However, we do want to find a way to make the use cases that we support more explicit. What I hear is that the auto-generated UI is important to this community; it's used by (I imagine) all of us for prototyping, and it's used by some of us for deployment into production. That said, I'm also pretty sure that many would-be developers *don't* see an automatically generated UI as being something they care about. Or, they may have negative preconceptions about the NO pattern. So for these people we should make it much more obvious that Isis' is also a great platform for REST. The other point that I'm hearing is we need to make it easy for would-be developers to verify our claims for themselves, hence the demo apps etc. Fully agreed on that. As you might have seen, Mike Burton is donating some server capacity for us, and I'm gonna work on that area. (Thanks, Mike!) ~~~ Alexander's slightly tongue-in-cheek question is funny, cos actually long-term I do think that Isis should lose its persistence layer. My view is that the fact that it can only be deployed on the handful of objectstores that we have implementations for counts against it. In time I would like to simplify the whole, rather complex, runtime/back-end, and plug into JDO or JPA or maybe have a really lightweight implementations around NoSQL (eg Mongo). Thus, Isis is (and will always be): (a) an extensible metamodel, predicated on the idea of behaviourally complete objects (b) it provides an abstraction over the domain object instances in some "runtime" (c) viewers of the domain object with respect to that metamodel. And of the viewers, there's a whole spectrum, of various customizability, but I think that the REST one deserves special prominence w.r.t finding a new audience, and the HTML viewer deserves special mention cos its currently the one that's the most robust. The others we de-emphasise until they have the additional work put into them to make it clear which use cases they target. Further thoughts/discussion very welcome. Cheers Dan On 23 November 2011 17:00, Alexander Krasnukhin <[email protected]>wrote: > You can remove persistent layer next and NO will magically disappear. > > On Wed, Nov 23, 2011 at 5:38 PM, Johan Andries <[email protected] > >wrote: > > > Oh. The *current* home page puts the n.o. pattern and the hexagonal > > architecture forward as the foundation of Apache Isis, but then again, > > that home page is about to be redesigned. > > > > But if you remove the n.o. focus, then what is left? Isn't Grails a > > better option if you do want to write a custom UI for everything (and > > keep it up to date all the time)? I guess I'm confused. > > > > Johan. > > > > On Wed, Nov 23, 2011 at 3:01 PM, Luis Bender > > <[email protected]> wrote: > > > Agreed, Johan. > > > > > > My perception may be wrong, but from the previous messages it seems > that > > > we're close to giving up the OOUI (and the NO pattern) instead of > > improving > > > it. > > > > > > Cheers, > > > - Bender > > > > > > -- > Regards, > Alexander >
