[Opensim-dev] Development models (was Re: The essence of "grid")

[Dahlia Trimble]

Personally I prefer OpenSim in standalone mode, and my preference is for it
to be a personal simulation server allowing multiple dissimilar clients to
attach and share a simulation; one which may deviate quite a but from the
normal SL experience. I realize I'm in a minority with this position
compared to other core developers, and as such I do a majority of my
development and testing in grid mode and with hypergrid using various
viewers, primarily those based on the LL viewer, but also using other
viewers not derived from the LL viewer.
Many of the users of OpenSim have their own ideas about how the platform
should evolve, and hypergrid appears to (at least anecdotally) be a popular
feature. It also requires substantial changes to the core architecture for
proper implementation, and bringing hypergrid and diva into core has allowed
her to make a substantial improvement to many parts of the code which deal
with standalone and grid operation in addition to hypergrid. I see no lack
of benefit to any of our users from bringing diva and hypergrid into core.

Anyway, last I checked, time still only moves forward so please continue to
offer suggestions for improvement and they will be considered :)



On Fri, Apr 17, 2009 at 9:14 AM, Mike Dickson <mike.dickson at hp.com> wrote:

> Justin, thanks for clarifying the process. And I certainly understand
> the interest in Hypergrid and the energy behind it. Charles your message
> was also helpful in highlighting to me what is at the center of my
> concern.  I agree the development process is somewhat chaotic and things
> get hacked in based on interest.  That's probably completely to be
> expected though it may not make for the best platform going forward.
>
> Using Hypergrid as an example,my preference would be to do it outside of
> core. So let me explain that.  Something like Hypergrid is going to
> require a different usage model from the original core (different
> protocols for "teleporting", now the exploration around inventory, etc).
> Rather than have the changes to handle that get introduced into core I'd
> have preferred to see something like an RFC that documents what is being
> proposed, and what "interfaces" need to be changed in order to
> accommodate the new use cases.  That RFC gets iterated and the
> interfaces evolved to make "hypergrid" possible as a pluggable module.
> Over time most likely the set of commonly used modules grows and you
> ultimately end up with a core framework and a "core" set of modules that
> define what the out of the box functionality of an installation is
> (standalone, hypergrid, what have you).
>
> The obvious problem with this approach is that it requires evolving the
> core framework which is not nearly as "sexy" as hacking in new features.
> I've done both approaches.  Certainly a cool demo can go a long way to
> sell a concept and often the change the framework process takes enough
> time that prototypes don't happen. It's more work to maintain a branched
> copy of core while you evolve your prototype into a set of changed
> interfaces that support it.  Personally I believe that more disciplined
> approach is the key to seeing OpenSim get to 1.0. And ultimately be a
> better platform for experimentation.
>
> So I like the concept of hypergrid.  I think prototypes like that need
> to exist if only to prove that the community is healthy. But I also
> believe that how the "framework" is defined and evolves is equally if
> not more important (to me at least).
>
> Just my 2 cents.
>
> Mike
>
> On Fri, 2009-04-17 at 15:35 +0000, Justin Clark-Casey wrote:
> > But I do have to also point out that OpenSim development is largely
> driven by the interest of the developers (since
> > there's no single company behind it).  If there's a lot of development
> interest behind Hypergrid then this is the
> > direction that's inevitably going to progress most.  If people coming
> along contributing code that enhances different
> > architectures, then development will also be driven in that direction.
>
>
> _______________________________________________
> Opensim-dev mailing list
> Opensim-dev at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-dev/attachments/20090417/13cacf9b/attachment-0001.html>