Open sourcing RISC OS
Aaron's take on a proposal from Castle Technology Ltd...
Recently Castle Technology Ltd (Castle) made an interesting announcement. The announcement came as part of an interview between Paul Beverley, the editor of Archive magazine, and Jack Lillingston, the Managing Director of Castle. To cut to the chase the announcement was that Castle were "very seriously considering making RISC OS open source", the principle being that a "a small royalty" would be charged for each copy used in order to fund handling of the main CVS tree.
This is very interesting news and sounds exceptionally positive, but, as they say, the devil's in the details. So lets take a slightly closer look at what Castle seem to have proposed. Firstly the proposal, so keenly seized on by a number of regular commentators on the RISC OS news sites, doesn't relate to RISC OS in it's entirety, just to elements of RISC OS. Peter Wild, one of Castle's directors added that "there are some commercially sensitive/valuable components that probably wouldn't be released". So, here's the first problem, what's being talked about isn't open sourcing RISC OS itself, but some parts of it.
That isn't in itself a problem, some components could be provided in binary only form, but lets dig a bit further. Peter also added that any licence for an open source RISC OS "would almost certainly be under a bespoke licence with appropriate terms for use, charges and feedback of sources." Well it certainly makes sense to have any changes made by a licence fed back to Castle, but what's a bespoke licence? What could the appropriate "terms for use" actually be? One of the interesting proposals that was made was to look at porting RISC OS to non ARM hardware, Peter's comment on that was," I really don't think this is practical, or will be legally allowed." This sounds rather odd, lets sum up where have go to so far...
So far this doesn't really sound like a normal type of open source to me. Still lets take a closer look at the possible charges...
Peter also went into some detail about charges, suggesting that there would be "a minimal royalty for personal use". On the surface this may sound reasonable, but again lets look a little deeper. A user would pay Castle for the sources, then be obliged to return any changes back to Castle and also be restricted on what products that could use RISC OS with for their own purposes. If they then wanted to produce a commercial product they would need to pay a "per unit royalty" based on a "structure for commercial use". I'm sorry, but this simply isn't how "open source" software licencing works.Unless the costs were very low, in the pence per item category, then Castle simply aren't going to get any takers, especially of the uses for RISC OS are restricted using some arbitrary rules. Any developer would need to ensure that not only were the price points reasonable but that their product fell into the definition of "Legal" as used by Castle.
So, having covered the money side of things lets look at what happens if a user makes changes to the sources.
Changing the sources...
As one would probably expect any changes to RISC OS made by an individual/company would need to be fed back into the main source tree. However there is a major potential problem here. There would be a requirement that sources be fed back, but sources would have to be fed back in such a way that any work done "can be re-licenced by Castle". So, if a commercial customer made changes to RISC OS for their own products they would be forced to hand these new sources back to Castle and then allow Castle to sell the new sources on to any third party. Can you imagine any commercial RISC OS developer working under these terms, when any sensitive commercial developments would have to be returned to Castle?
This is especially ironic given that a few years ago Castle's insisted that work done to support their RISC OS 4 machines shouldn't be re-licenced by RISCOS Ltd to other third parties. This was one of the sticking points that helped kill off any hope for the RiscStation Evolution.
So to sum this up, Castle would own rights to any changes made to the RISC OS sources.
As we have seen Peter has already said that there would be restrictions on what a licencee could do with RISC OS. Firstly any form of non ARM hardware wouldn't be allowed, so you can forget anyone attempting to port RISC OS to another processor/platform. So what would be allowed? Well presumably a potential licensee would have to tell Castle exactly what they are up to before starting work. So not only could any commercially sensitive code be re-sold by Castle, but they would know exactly what was being worked on, so they could line up a new customer for the same product before the sources were even passed back.
In a wider commercial context it would be potential suicide for any commercial company to take on RISC OS under these terms. That of course assumes that commercial licencing is what Castle's "open source" plans are about. However I don't think that commercial licencing has got anything to do with it...
So what's really going on...
I think that there are two issues involved here, the first is PR.
Castle's announcement was met with a lot of positive comments, at least until Peter Wild starting spilling the beans. This does raise an interesting point. Were Peters' comments deliberately designed to cause the spin to backfire? Given what I have been told about the relationship between Peter and the other Castle directors I have a sneaking suspicion (although no direct evidence) that this might be the case. So, my view is that the first issue was to get some positive PR easily and quickly. This could well be very important given that Castle could do with some good news at present. Of course it's mainly backfired with only a few Castle fans, who can't see the commercial realities, remaining positive.
The second, and I think possibly more important issue relates to the continuing development of RISC OS. As RISCWorld readers will know Castle no longer have any full time development staff. John Ballance from Castle is continuing work on RISC OS, but releases are slow in coming out. Suppose Castle could get some members of the RISC OS community to continue development of RISC OS and also get the community to pay for the privilege into the bargain? So Castle would get the work done for "free", would charge each developer "a minimal royalty for personal use", presumably including getting access to the sources in the first place, and would also have all the rights to the work that's being done. Now that does sound like a sensible plan, doesn't it?
If one assumes that what Peter Wild has said is correct, and after all he is a director of Castle, then the above does sound very likely, as the terms apparently being offered would seem to be not greatly attractive in a commercial environment. Anyway, lets sum up what's being proposed so far one again:
As you can see this doesn't look attractive, but there are two more absolute dead stops in the way...
So who really owns RISC OS?
This isn't a straightforward question. Peter Wild has in the past claimed that Castle own RISC OS "lock,stock and barrel." Complete rubbish. We have covered this in the past in RISCWorld, but it's worth going over it again quickly. Believe it or not Acorn never "owned" all of RISC OS in the first place. Although Acorn owned some components of RISC OS, since they were developed in house, a large number of other components were actually licenced from third parties. In some cases this was from Acorn employees, in some cases from third party companies. Although Acorn had rights to use these in binary form, they were never owned by Acorn. This didn't present a problem when Acorn licenced RISC OS to third parties, such as RISCOS Ltd, because Acorn had the rights to make such licences. What Acorn never owned was clear title to RISC OS.
When RISC OS was passed to Pace it was done so on the understanding that although binary use was covered the actual ownership of much of RISC OS was a grey area. Before Pace "sold" the "RISC OS technology" to Castle they attempted to sort through it to see what they actually owned. The list of items for which Pace had a clear title was surprisingly small. Most items ended up in either the "we probably own it", or the "err...don't know" box. Castle "purchased" RISC OS based on this understanding. Here we find the other dead stop, Castle's purchase agreement with Pace.
After some hard detective work RISCWorld has tracked down official Pace documentation relating to the Pace sale of RISC OS and Castle's "purchase" of RISC OS. The paperwork makes very interesting reading, but has one clear point to make. Castle have no rights to open source RISC OS, indeed the purchase actually forbids Castle making RISC OS open source. This may well explain why Castle become so agitated when RISCOS Ltd made some parts of !Printers open source on SourceForge. If you are interested it's still available from here.
So why can't RISC OS be open sourced? Because Castle don't have the right to do so. They don't have the right because Pace never had the right to do so. Pace never had the right because Acorn never did. Admittedly Castle have recognised this problem and Pete Wild spent some time trying to track down the actual ownership of the IPR (Intellectual Property Rights) with, what I have been told was, limited success.
The only items that could, perhaps, be made open source are those that Castle have 100% title to, even then there is the little matter of the sale/purchase agreement with Pace to deal with.
As it stands RISC OS can't be made open source. But then, if you read what has been said by Castle directors this isn't the plan anyway. The plan was to make the sources available under licence for a fee per use. Then get hold of the full rights to any changes made by the licencee so they can be licenced on by Castle. The plan is simple, get RISC OS developed further without spending any money. This may be a very good proposal, but I'm sorry, it isn't open source.