Was this week all about Unity, mobile or what?

This week has seen several developments relating to Unity3D although the image above comes from the SL/OpenSim viewer Radegast of which more anon.

Firstly Tipodean announced their OpenSim-to-Unity3D conversion service and provided some nice demos. The benefit is that, depending on the mode, the sim runs either serverless in single-user mode or peer-to-peer in multi-user mode, albeit that some avatar management might be needed (see Unifier later). Moreover, the sim should run on the two major mobile platforms, iOS and Android, as well as the web (with a small install). This answers two important questions: scaling (OpenSim equivalents might be sim-on-a-stick and Intel Distributed Scene Graph) and mobile (no mobile clients with inworld scene graphics as yet).

It's perhaps worth mentioning that Rezzable demonstrated something similar in late 2010 but as yet there is no product or market. Although it only bears a conceptual reference to OpenSim, ReactionGrid have been busy developing their Jibe Unity3D product and features such as mobile, inworld building and hypergrid-style teleporting appear on their roadmap. Just this week Second Places announced their Unifier product for managing multi-avatar Unity3D spaces.

Having been early on the scene, Rezzable have commented at some length on the relative strengths of the OpenSim and Unity3D platforms and the ways in which they might complement one another. Tipodean have made the case for Unity in terms of faster onboarding while Rezzable like the idea of a taster that might eventually lead to download of the full viewer. I think everyone agrees that a simpler first hour experience is desirable as evidenced by the new Basic mode in the SL Viewer 2. Maria Korolov followed this up more recently with a specific comparison of OpenSim and Jibe.

For me (as a non-expert), it seems that OpenSim is preferred if you want low-cost regions in which avatars can rez and interact with content while Unity3D provides a more contained experience with a somewhat smaller install and the ability to deploy on the web as well as mobile platforms. Depending on your application, either might be appropriate and the Tipodean product is useful in so far as it bridges the two platforms. At the moment it doesn't convert LSL scripts to C# nor, I suspect, avatars so there may be additional customisation to be costed post-conversion. Those who have worked to develop their SL/OpenSim skills may also feel a little disenfranchised though their students will know no different in many cases and the early adopters will at least have some experience with the pedagogy.

The spectre at the feast is arguably 3Di who had OpenSim running under Android back in 2010 using Flash conversion but who have now stopped OpenSim development to focus on 2.5D environments.

I also feel a little that way as my new, low-end Android tablet will not run the Unity3D demo games on Android Market so I don't expect to be able to run Unity3D sims except in a browser. That means, alas, that I can't comment on likely usability issues derived from the smaller form factor, for example. I therefore simply observe that there might be other 3D solutions for mobile Android, albeit of lower graphics quality.

Which brings me back to the Radegast viewer widely used on PCs lacking SL-standard graphics cards. Formerly a largely text-based viewer, the development version has now sprouted a 3D option in addition to an object inspector. While not approaching the quality of, say, Imprudence and presently rendering avatars with the default mesh (aka Ruth), the scene can be explored using alt-zoom camera techniques and objects are rezzed in response to touch-activated scripts. It looks like sitting/standing, take/return/delete and rezzing from inventory will all be supported.

Kitely have been promoting Emscripten as a means to convert a standard viewer to JavaScript so that one might have support for HTML 5 and truly plugin-free performance. Radegast would seem to my superficial inspection to be an interesting starting point, albeit that my tablet still could not render the scene as it lacks WebGL support. Woe is me unless Google does some significant magic.

Which brings us to the lowest acceptable level of performance, perhaps, viz something that would take a panoramic series of images, stitch them together, upload the file to a website and then provide a basic viewer that allowed you to spin the image to the desired location, navigate via coordinates/hypergrid addresses and chat. Rinse and repeat at, say, 60 sec intervals. Something like the SL-based Peek360, for example, which runs fine in the default browser on my tablet but is only updated through manual intervention and does not guarantee punctual delivery. Of course, we need something that is not Flash-based (for iOS) and we need additional OSSL support or some kind of libomv-based service as is the case, I suspect, with Peek360. I haven't done an exhaustive search for viewers but this one, for example, scrolls satisfactorily in the default browser. Not futuristic, not elegant and a service that would cost something to run. For some purposes, however, it might suffice "for the rest of us".

Basically, it's all about having options. Options are good.