Open source e-learning technology hits prime time
So stand by for the inevitable backlash. But seriously: the relation of open source to open standards was the topic of the opening debate of the big alt-i-lab e-learning technology interoperability do, the open sourcing of LAMS was announced there, the increasingly popular Open Source ePortfolio Initiative (OSPI) held a conference right before it, open source reference implementations were mentioned by nearly everyone there, most of the hotly debated service oriented approaches to MLE design rely wholy or in part on open source components, and even Microsoft announced its inimitable take on the phenomenon by launching a 'shared source' webservices kit for Class server.
Because alt-i-lab brings together a lot of different people with an interest in e-learning technology interoperability, it can be a fair, if unscientific, barometer of trends in the e-learning tech domain in general. The sheer diversity of tech buyers, vendors, developers and spec experts means that the trends can vary from worthy but vague agreement, via polite agreements to disagree, to outright religious wars and even solid collaborative work.
In that volatile mix, open source solutions to e-learning used to be seen as quite wacky and out there; its developers wouldn't shout too loudly about it for fear of upsetting "the vendors". Even some of the main intended beneficiaries —the technology users— would look with wariness at these here-today-gone-tomorrow hippies with their quirky tools. With the very high profile of open source solutions today, that era is clearly coming to an end.
The open source and open standards debate was, in that sense, evidence of a shift in gear: at issue were not whether open or closed source solutions were better at supporting standards or anything else, never mind the earlier necessity of making clear that open source and open standards are not the same thing. Instead, Brad Wheeler from the open source Sakai project and Chris Vento from commercial vendor WebCT focussed on how the two software development models relate to each other and to open standards.
Brad argued that the relation boils down to the question who will get most value out of standards: the vendors or the buyers? Implication: mostly the buyers. Vendors have to listen to their customers, but would rather sell as many of their own widgets as possible in perpetuity. Brad pointed to the tension between generic functionality and unique added value too. If all systems do all their functionality in exactly the same way, what reason is there to choose one over the other? User-led development under open source licences means that vendors are forced to give standards support higher priority. Both because they'll be less likely to supply all the widgets, and because their products have to remain competitive on all features, including interoperability.
Chris's reposte was remarkable for its recognition of the fact that institutions are likely to have mixtures of open and closed source solutions. In each area, buyers will choose what suits them best, whatever the licence conditions are. Hence the need for common, modular integration frameworks, with open standards to define the crucial interoperability points.
The debate that followed quickly shifted attention to the second major open source issue: reference implementations. The idea here is that for new or emerging specifications, it is best to start building a basic open source implementation, both to make sure the spec actually works, and to give everyone the same idea of how the spec should be interpreted. Both commercial and open source developers can collectively improve the reference implementation, and incorporate it in their products.
CETIS' Bill Olivier wondered what it is about one open source implementation that makes it a reference implementation, to which Brad replied that it is the one that first solves a particular problem in a 'good enough' way. Carnegie-Mellon's Dan Rehak and CETIS' Scott Wilson reminded everyone that complete applications are seldom effective as reference implementations. The latter are better left as code libraries that fulfill a specific interoperability painpoint.
The point is illustrated by both the soon to be open sourced Learning Activity Management System (LAMS) and the Open Source ePortfolio Initiative (OSPI): both are complete applications that aim to converge with the relevant standards (IMS Learning Design and IMS ePortfolio, respectively), but were not designed to implement them from the start, never mind be a reference implementation of them at this point.
In the area of service oriented approaches to Managed Learning Environment (MLE) design, the same point is illustrated from the other side. Though the Open Knowledge Initiative (OKI) and the E-Learning Framework (ELF) may release the crucial interoperability interfaces, and reference implementations of the interfaces, as open source, but neither assumes that the systems they connect will do the same. OKI Open Service Interface Definitions (OSIDs) have fledgling implementations in commercial tools such as WebCT's Vista and Giunti's LearneXact, just like Icodeon's new Simple Sequencing engine will be exposed as a webservice in the ELF.
Though many a debate can and will be held about whether Microsoft's newly liberalised 'shared source' licence can be compared to regular open source licences such as the GPL or MIT's, even there it is recognised that more people should be able to look at, modify and share the service interfaces. Even if the platform that runs them, Class server, is still very much closed source commercial software. We'll have a closer look at that solution soon.
So what about that backlash against open source? It will be truly massive, if enough people think that slapping a GPL on a piece of software will instantly transform it into a slicing, dicing, kitchen sink included, world beating reference implementation that solves everyone's problems...
Brad Wheeler's powerpoint presentation for Alt-I-Lab 2004 is available from the IMS website, as is Chris Vento's.