LWN.net Weekly Edition for February 16, 2006

Toward a free metaverse

Second Life is one of many multiplayer role-playing games springing up on the net. Unlike some others, Second Life gives its players a significant amount of freedom to create their virtual world; players also get ownership rights to their creations. The Second Life developers have made an effort to strengthen the ties between the virtual and real worlds; the idea of making a living in the Second Life "metaverse" is heavily promoted. Various "real life" personalities - Lawrence Lessig, for example - have made high-profile appearances there.

Like many such services, Second Life has built itself an infrastructure heavily based on Linux and other free software tools. Also like many such services, Second Life has returned that favor by providing clients for proprietary operating systems, but not for Linux. Until recently, that is. There is now an alpha-test Linux client available for free download. It is still very much a proprietary client - no source, x86-only, etc. But it is a start, at least.

Your editor would, of course, rather be reading memory management patches or following the interminable "bait Joerg Schilling" festival on linux-kernel. But journalistic ethics required that time be taken out from such rewarding activities to see how this new client works. LWN readers would expect no less.

Alas, no joy was to be found in that direction. This client, it seems, requires a 3D-capable graphics card. It also requires that the proprietary driver be installed for said card. Your editor is willing to make many sacrifices for the cause, but jumping into the world of binary-only kernel modules was pushing things a little too far. So no Second Life; there was real work to get done anyway.

An important thought came out of this exercise anyway. We, in the Linux community, will certainly want to be able to participate in this sort of virtual universe in the future. These worlds will only get more realistic, engaging, and compelling. They will host a growing number of real-world meetings and events. Even if we, personally, have no particular yearnings for a second life in the virtual world, we may well end up going there just for a chance to visit our children. So it is important that Linux users are not excluded from this sort of experience.

Unfortunately, the lack of free drivers for contemporary video hardware threatens to exclude us from that experience. Even those of us who see no need for a 3D version of vi (emacs, of course, will have a full 3D Lisp mode) may have the occasional desire to dress up as some sort of furry animal and commune with the virtual world natives. Much of what will be interesting in the future of computing will involve increasingly realistic interfaces - and that will require good graphics support. If Linux does not provide that support, people will use something else.

Assuming we will eventually get past that issue, there is another, more important question to be answered: why, exactly, should we build our virtual worlds on somebody else's substrate? Even if we "own" our creations, they run on somebody else's server (which they can unplug at any time), uses their currency (which they can degrade at any time), and is subject to their rules (which they can change at any time). A virtual world which is not free is, well, not free.

Instead, the creation of a true "metaverse" should be a project which is a natural for the free software community. A decent set of open protocols and libraries should make it possible for interested people to set up their own neighborhoods on their own servers and tie them all together into a distributed - but integrated - whole. The net was built on free software, and it has served as a platform for no end of interesting developments. If we build our virtual worlds on free software as well, people will, beyond doubt, create environments beyond our wildest imagination. It is hard to see why we would want it any other way.

There are some virtual world projects out there. MUPPETS is an academic project with an educational focus; its last release was last July. MUPPETS appears to be a Windows-only application, however. The Croquet project looks like it is oriented toward people who want to get some real work done, but it looks like it could be put to wider uses. The Open Source Metaverse Project is an attempt to make something very much like a free Second Life. This project appears to have stalled, however; there is still some life in its forums, but the last development release was in August, 2004. Solipsis is an attempt to create a true, distributed virtual world, but it is at a very early stage. Interverse has some nice screen shots, but the project appears to have come to a halt. Verse is a 3D-oriented network protocol associated with the Blender project; it looks like it could be a useful component. The Virtual Object System is a collection of projects around the 3D, virtual reality theme; it released version 0.23.0-pre1 in January. And so on.

So there's a number of projects out there, but it is not clear that any of them have truly reached a critical mass. One would think that such an inherently fun project would attract more developers. Evidently free software developers have other itches to scratch. So we may find ourselves, in the future, building our virtual worlds on non-free platforms and hoping that the Second Life folks live up to their hints that they might open up their protocols - in 2010.

Comments (14 posted)

On the dual-license model

When people go looking for successful free software business models, the dual-license approach tends to turn up near the top of the list. With this model, a company releases a software component under a copyleft-style license so that all may make use of it. This company also offers the same software (or, perhaps, an enhanced version of it) under a paid, commercial license, allowing other companies to incorporate it into their products without the need to make their own code available. This model will clearly be most successful for software which works as a building block for larger systems. The dual-license model has been employed by companies like MySQL AB, Trolltech, FSMLabs, Sleepycat Software, and others.

The dual-license model can look like the best of both worlds. The free software community gets high-quality, supported code - and often good documentation as well. Developers get paid for working on that code. The company which makes all this happen gets to stay in business. That company's customers get (1) the use of the code in their proprietary projects, and (2) an immediate indication of what it is costing them to keep their own code non-free.

This model is not suitable for every project, and it is not without its disadvantages. One of the strongest of those, perhaps, is the disincentive it presents to potential contributors. A dual-license company can only accept contributions which it will be able to sell under its commercial license; in practice, that implies copyright assignments or some other form of explicit permission from each contributor. Some developers are happy to contribute code under such conditions - those contributions improve the free version of the package, and the developer still probably gets much more back than he or she ever could contribute. But others are less interested in contributing code which can be taken proprietary for somebody else's exclusive commercial benefit.

Another potential snag in the dual-license model was highlighted this week when Oracle announced its acquisition of Sleepycat Software. Like Innosoft (acquired by Oracle last year), Sleepycat provides a transactional engine for MySQL's database offerings. MySQL gets that code under a dual-license arrangement which, in turn, allows MySQL to include it in its own dual-license products. The result is that Oracle now controls two important components shipped by MySQL.

Sleepycat CEO Mike Olson says that neither the free software community nor Sleepycat's commercial customers should be concerned about this acquisition. But Mr. Olson spoke a little differently after the Innosoft acquisition:

(Thanks to Jim Thompson for the pointer to that article).

It is worth noting that neither acquisition can do immediate harm to the free software community. The code which was released under a free license remains free and cannot be taken back. The worst-case scenario would appear to be that developers could be taken off the projects, slowing or stopping the development of that code.

The situation might be a little more perilous for MySQL AB, however, and its customers as well. Oracle is now in a position to change the licensing terms for both database backends, or even to make them unavailable for dual-licensing altogether. And that points out an important aspect of the dual-licensing model: if you buy into the proprietary side of dual-licensed software, you are very much in the proprietary software world. And, at that point, you can be impacted by policy changes by your supplier - or by their suppliers as well.

Buying proprietary access to dual-licensed software may still be the best path for many companies. It can enable the use of high-quality, community-reviewed software at a reasonable price. But dual-licensed software should not be seen as free software with some commercially inconvenient strings removed. It is proprietary software, with all the risks that come with the proprietary model.

Comments (18 posted)

Another analyst TCO report

Yet another analyst report comparing the costs of running Linux and Windows networks has been released. The report was funded by a corporation with a clear interest in the outcome, but, of course, the authors claim to have done entirely independent work. It features data collected from a number of different companies (the way these companies were selected is not disclosed) and from "self-selected" respondents to a web survey. Information on the availability and cost of administrators was obtained from "a cursory survey of resumes" from online job boards. Surprisingly enough, the report is strongly favorable to the company which sponsored it.

The Linux community, once again, has come together to debunk the findings in this survey. Well, actually, maybe not. This report was sponsored by Levanta and OSDL, and is unequivocally favorable to Linux.

Those who are interested in the details are encouraged to look at the press release, the executive summary, or the full, 21-page, pie-chart-stuff report [PDF]. In essence, however, it says this: Linux systems are cheaper to purchase and install, cheaper and more reliable to administer, and more secure than the alternatives. Linux administration staff can be had cheaply, and is in plentiful supply. Oh, and if those administrators are well equipped with "sophisticated administration tools," such as those sold by, say, Levanta, they'll be even more efficient.

Much of what is found on these glossy pages corresponds to the experience of those of us who have managed large networks of systems. A Linux administrator really can manage more systems than a Windows administrator. But the sad fact, which not all in the community seem to want to recognize, is that this report is the same sort of subjective analyst recycle bin fodder that the proprietary software companies crank out. We should not invest it with a higher level of credibility than the other offerings in its genre.

It is worth noting that this report appears to have had the desired initial effect. The technical press has dutifully carried the "Linux is cheaper" news. Presumably, the pointy-haired bosses who are held to be impressed by these reports will be suitably influenced. It seems that these analyst reports are simply part of how this game is played. People who are trying to get some real work done on a Linux platform need a stack of glossy paper to justify their decisions to certain levels of management. The other side is producing a long stream of these reports; if the Linux side has no reports of its own, it looks like it has no answer at all. So it may be a good thing that somebody is going to the effort of producing all this paper. But we shouldn't make the mistake of believing that reports like this one prove anything.

Comments (4 posted)

Page editor: Jonathan Corbet

Inside this week's LWN.net Weekly Edition

• Security: Nmap 4.0; New vulnerabilities in gnutls, heimdal, libpng, postgresql, java...
• Kernel: EXPORT_SYMBOL_GPL_FUTURE(); Tweaks to madvise() and posix_fadvise(); A new approach to robust futexes.
• Distributions: Building a Custom Live CD with Linux-Live
• Development: Cracking Passwords with John the Ripper, new versions of Linux-HA, PostgreSQL, LAT, Oggz, Dada Mail, OpenSSH, Sussen, mod_python, TWiki, CLAM, GNOME, GARNOME, GnuCash, ScummVM, MH-E, Fastbreeder, Azureus, BioImageXD, Gourmet, Roundup.
• Press: UK holds Microsoft security talks, Is ODF an Open Standard, how Microsoft sabotages Linux, the Meshnode router, using Ethereal, Remote X with FreeNX, FSF Launches Gnash.
• Announcements: BEA donates Kodo, new Linux/Windows TCO, MySQL AB secures $18.5M, Oracle buys Sleepycat, EFF Challenges Clear Channel, GPLv3 Update #2, aKademy-es 2006, European Common Lisp, Freedom-to-Connect, FUDCon CFP, Libre Graphics Meeting.

Next page: Security>>