Technological improvements in Second Life? You bet!

Theatrically reprinted from hot chick Gwynneth Llewellyn 🙂

October 28, 2010 – in Homepage, SL Technology by: Gwyneth Llewelyn

When Philip “Linden” Rosedale announced, once again, that he’d be leaving Linden Lab and was actively searching for a new CEO, leaving Bob Komin to replace him temporarily, some of my friends saw this as a clear sign that Linden Lab was shutting down Second Life® — and point to the ToS changes which allow LL to shut SL down without paying residents anything, refund them, or compensate them in any way.

Others immediately thought that Linden Lab would go IPO, or, well, had already been bought, and that the new owners would soon nominate a new CEO. And that the ever-growing list of “bad news” — one or two measures announced every week that will only hurt residents — would just justify the upcoming impending doom.

I personally don’t believe any of those three possibilities, because they make no sense in the light of the under-the-hood work that LL has been doing the past few weeks. In fact, even in spite of many glitches, LL is actively developing the core technology at a pace that we haven’t seen since 2006. And there might be two good reasons for it: they’re trying to match innovation and speed of development of the third-party viewers (a shameful thorn on LL’s innovative stance; innovation has been coming mostly from the TPVs, not from the ‘Lab), and they’re taking the OpenSimulator exodus as more serious than before (there is a lot to be said about this). Former Lindens are now joining the TPV developer community (like Qarl “ex-Linden” Fizz, releasing his amazing prim-alignment tool which will only be available on TPVs), or becoming part of prominent OpenSim grid operators (like Pathfinder, the new Director of Community Development for ReactionGrid, a really logical role for someone who always was very close to the educator community currently migrating to the academic-friendly ReactionGrid, which has Microsoft and IBM as partners).

So, one would expect LL to become more serious about infrastructure stabilisation and innovative new technologies — while trying to make at least a serious effort to minimise some of the more hated features of SL here and there.

And innovation there is. Jack-of-All-Trades Linden announced the new SL 2.3 Beta — just a week after SL 2.2 became the official SL viewer — and which has a lot of niftiness, as well as some more polemic features. Gone is the “jumping screen” every time you clicked to open the sidebar. And, of course, this was the pretext to launch Display Names on the main grid too: here goes the picture that I had always wanted to do, cloning and impersonating my own self:


Fun besides (I’ve already addressed the issues why this is not such a good idea), there are a lot of under-the-hood developments. For starters, one thing that always frustrated me with Viewer 2 is that “Sort by most recent” (e.g. the old “sort by date”) was always broken since the very first release of SL 2.0 (where it worked). 2.3 Beta finally deals with this issue, and I’m surprised that it wasn’t fixed before — allegedly, it only affected some Mac users, including yours truly, so it wasn’t deemed to be important enough.

But the most fantastic change for me was fixing, once and for all, the many issues about Alpha Textures. Some preliminary work has been popping up in the latest betas and “development versions” (I have now five LL viewers installed on my disk!). This time, they got it right. The trick seems to be enabling an option on the Develop Menu, under Rendering, where it says “Automatic Alpha Masks (non-deferred)”.

Excited, I went to the sim with the most intense alpha texture issues that I know: Neufreistadt. The city is built at the cloud level, so SL has to deal with a lot of alpha textures: fog and clouds, which are constantly in motion across buildings — low-prim buildings with windows made by alpha textures. Add a bit of glow to make things even worse, and you get pretty much this:


This is how Neufreistadt has always looked like for us, at least since October 2004. No TPV renders it right, either (except Kirstens Viewer, which is based on the new alpha rendering pipeline).

Here is how it looks now:


I mean, this is just… wow! At a flick of a checkbox, all problems were suddenly solved. Even “legacy fog” looks nice for a change! Nobody has ever seen this sim looking so good As a nice side effect, invisiprims on shoes also render correctly on top of alpha’ed shoes. So, LL, whatever magic you’re doing, you’ve finally hit gold!

Also, for the first time in a long time, SL 2.3 Beta beats Imprudence in raw performance (that is, FPS) using the same settings (or as closely possible to the same settings, since the 1.X settings are different than the 2.X ones). There is a lot of trickery going under-the-hood. I haven’t tried out Phoenix yet, which is supposed to be far faster than Imprudence or any other TPV, but… still, LL did some homework. Things are getting better. It still takes some time to get used to the 2.X UI, of course, but for the casual user (I’m not a builder!), 2.3 Beta is much closer to what 2.X ought to have been since its release.

Alas, the improvements don’t stop at the viewer level. For the past few months, most people have missed an important change on the way LL now releases server code to the grid. I wish I had found the announcement that explains this in detail, but, currently, the main grid is split into three “channels”. According to my understanding, this means that Linden Lab can effectively deploy three “prototype” simulation servers, on the main grid (besides the ones on the Preview Grid), and have them all “live” simultaneously, with real data and real avatars. Each “channel” prototype is deployed on about 10% of the grid (meaning that 70% run the “main” release, and 10% each a prototype release). This is why it has become increasingly common to get that message saying “The sim you teleported to has a different simulator version”. Most residents don’t even bother, and have no clue that they’re actually helping to test out three different prototypes, one of which will become the “main” release after a while (that is, the one that gets deployed across most — but not all — sims).

Each “protptype” addresses different things in isolation (a released version tends to affect a lot of different areas of SL simultanoeusly, e.g. physics, avatar tracking, server-client communication, and so forth), thus allowing LL to measure the impact of each bug fix/improvement/feature without interfering with the rest of the code, and also to establish a benchmark in contrast to the “main” version. It’s a very clever way to deploy new code, new features, improvements, radical new changes, and have them all “live” for people to try out. For instance, although officially Display Names are supposed only to be available in January, you can use the Viewer 2.3 Beta and log in to the 10% of the sims that already have that option turned on, and watch Display Names in action (that was, in fact, how I took the first picture: it’s on the main grid, not the preview grid!).

And LL has been busy. Take a look at the Wiki pages for the three channels, and look at the calendar for deployments on Every few days, there is a new deployment of at least one of the new channel prototypes, but sometimes all get deployed at the same time. Estate Owners can even select a channel to subscribe to — so they will be able, if they wish, to test out a prototype version instead of the regular one.

If anything, the pace of development is increasing — not slackening down. Of course we all expect development to be even faster and faster, but there are limits to how much LL can do. Nevertheless, they’re not idly waiting until “something” happens — The End of Second Life As We Know It, an IPO, a buy-out, or anything dramatic like that.

Obviously, if you’re paranoid, you’d say that this is exactly the behaviour to be expected from a company wishing to improve their image in order to go public or be sold. I have my doubts. This past year has shown how terrible LL can be with PR — we never got so many “bad news for residents” in such a little time. If they wished to express a good corporate image, based on excellent relationships with their customers, they couldn’t have failed more. Their communications have been amateurish and devastated by the media. Every single bit of “bad news” could have been presented in a different way, handled with more care, showing that LL doesn’t, after all, hate their residents by harming them. I find it very hard to believe that this is a strategy to “get sold” or “go public”; and if they’re planning to shut down anyway, what’s the point of improving the viewer, the simulators, and the overall infrastructure if they don’t plan to be around for long? It doesn’t make any sense.

Granted, LL is not famous for making sense, but I still maintain that all the above shows that LL is really taking notice of what the competition is doing — and the competition are TPVs and OpenSim, not IMVU or Blue Mars, while Unity3D-based VWs might start to become the primary choice for education/business in a few years, if LL doesn’t do anything to stop that from happening (meshes are the first good step; a viewer on a Web browser is the next). But Philip, before he left, clearly pointed out that the focus is going to be on the residential market.