RISC OS is open for business!

The number of commenters on those other news sites who seem to think that the source is now available for the first time is really depressing :-(

https://www.theregister.co.uk/2018/10/23/risc_os_open_source/

“Since the start of open-sourcing RISC OS there has been talk about the whole toolchain issue time and again.” – RISC OS has never been open-source… Hopefully, this Saturday, will mark the beginning of that.

It will be interesting to see who forks it first; If I was ROOL, my first move would be, fork, relicense to GPLv3, upload to github. Otherwise the ‘control’ of the code will quickly be lost :)

Personally I think there is a lot of interest in RISC OS… but for backporting and running on real Archimedes/RISC PC machines (or getting a close approximation of one on a Pi to play original games) – which has never been the direction the few core developers have taken – hence the total lack of wider interest. This unending dream of it suddenly becoming ‘commercial’ is totally lost on me. Still that’s just my opinion; we’ll see :)

there once was an LCC port available – not sure where that went

A while ago (i.e. probably over a decade ago) I did update LCC to be able to produce module code – can’t remember whether it got released though. Possibly it also needs fixing to produce 32bit output. But the bigger issue is likely to be that the version we currently have is unlikely to support C99, or some Norcroftisms like the inline assembler.

I’d say that a modern compiler like LLVM/clang would be a better place to start, but then you have the problem that modern compilers like to drop support for old architectures and ABIs, whereas RISC OS doesn’t. So there’ll be some degree of maintenance effort to make sure that things stay working. Although maybe that won’t be so bad, if the developer pool grows big enough to help push compiler changes upstream and make sure RISC OS is treated as a first-class target. But for that to happen we’ll need developers who are willing to get stuck in with hard tasks, and it remains to be seen whether relicensing the OS will help us in attracting those kinds of people.

“Since the start of open-sourcing RISC OS there has been talk about the whole toolchain issue time and again.” – RISC OS has never been open-source…

Oh please, not again…the source was open, so it was open source. It just was not Open Source according to the OSI definition.

Whether relicencing changes anything – I remain sceptical. Especially because of…

Personally I think there is a lot of interest in RISC OS… but for backporting and running on real Archimedes/RISC PC machines

…strange ideas like this (the relicensing does not change a bit for anyone trying to backport it to Archimedes or Risc PC machines (and why backport it to Risc PC machines anyway, there is a perfectly working RISC OS 5 version for the Risc PC and RPC emulators as well!) seem to still exist…or…

(or getting a close approximation of one on a Pi to play original games)

…which is nearly exactly what ADFFS quite successfully provides, as well as ArchiEmu…and why did the Castle license prohibit anyone from writing the perfect emulator?

– which has never been the direction the few core developers have taken – hence the total lack of wider interest.

…everything was open for developers already before the relicensing now announced, so anyone interested could have taken any preferred direction. I never noticed even the slightest (serious!) interest in going in any alternative direction, mainly because most of the “interest” was all talk and no action. Getting your hands dirty seems to be highly unpopular these days.

This unending dream of it suddenly becoming ‘commercial’ is totally lost on me. Still that’s just my opinion; we’ll see :)

“Commercial use” is nearly the only thing that imediately profits from the now-announced relicensing, which makes it even more difficult to understand the point you are trying to make.

I’d say that a modern compiler like LLVM/clang would be a better place to start, but then you have the problem that modern compilers like to drop support for old architectures and ABIs, whereas RISC OS doesn’t.

I don’t know LLVM in detail, but my idea would be to have an LLVM interpreter/VM as well as an LLVM-to-whatever-ARM-variant-RISC-OS-needs “translator”.

Clearly not an easy task, and to be honest I cannot see that happening – the few (two?) part-time developers of GCC/GCCSDK really struggle to keep the stuff half-way up-to-date, so dreaming about yet another compiler infrastructure project for RISC OS seems to be very far-fetched.

“But it is a car” – just not by any accepted definition of car i.e. it’s a unicycle. Seriously… as for the rest; well, if that’s what you want to believe, it’s a free country. Once (if) RISC OS really goes open-source; then we will see.

If I was ROOL, my first move would be, fork, relicense to GPLv3, upload to github. Otherwise the ‘control’ of the code will quickly be lost :)

I would imagine some would prefer Gitlab, as Github is owned by Microsoft. MS have been making attempts in recent years to atone for their past; however, that hasn’t stopped some repos from being moved from Github to Gitlab. Bitbucket is another alternative – Atlassian are an Australian company – that part of the world did have an Acorn presence in the dim and distant past…

RISC OS has never been open-source…

Funny. I have the source. I built my own ROM. I didn’t pay anybody for licencing deals to do that.

RISC OS may not have been “open source” in your definition, but it’s considerably more “open” than the GPL, a supposedly open licence.

just not by any accepted definition of car

Citation needed for the “any accepted definition” assertion. Personally, I think Castle’s “Shared Source” license fits e.g. the Wikipedia definition of “Open-source software” quite well.

But the redefinition of once-clear wording (like “free software”) is common in IT (and everywhere else), so no surprise here. Find two people who agree on the finer details of what is really meant by “free software” and/or “open source”…

“RISC OS may not have been “open source” in your definition, but it’s considerably more “open” than the GPL, a supposedly open licence.” – it’s not my definition – it’s the widely accepted definition of the term… this argument raged 10+ years ago, it’s now settled. So, if you really believe Castle’s licensing was more ‘open’ than GPL I can only suggest that perhaps you should learn a little more about licensing; since pretty much the whole software development universe disagrees. But, don’t take my word for it, post your observation over on stack overflow or some similar place and see what others think.

since pretty much the whole software development universe disagrees.

Bovine excrement.

If the whole software development universe agreed, there would only need to be one licence. There would be no arguments. There would be no “let the lawyers sorry it out” crap in the FAQ, and the Linux kernel would be GPL 3 instead of sticking to v2 because v3 is seen as too restrictive.
The entire world, more likely, doesn’t want to annoy the guy that pretty much kick-started the free software movement. However GPL does not play well with others, and each version is more restrictive than the one before.

We’ve had this argument before. At length. When it’s possible to take GPL code and combine it with other stuff to make a finished whole without implicating any specific licence to the other parts, then we can talk about what “open” really means. Until then, keep drinking the Stallman Kool-Aid.

Putting the old “one ring to rule them all” arguments aside, any other thoughts, questions or spots of this in the media?

https://www.theinquirer.net/inquirer/news/3064969/acorn-computers-risc-os-finally-goes-open-source

Dragging this back on topic, one of the nice things about the relicensing is that it should make redistributing projects that build on parts of RISC OS a lot easier. I can think of two straight away: the Linux port and R2.

But, don’t take my word for it, post your observation over on stack overflow or some similar place and see what others think.

You’d have to be very careful in wording this, as usually licensing questions are quickly closed as off-topic on StackOverflow.

Amusingly, you are not allowed to post GPL-licensed source code on StackOverflow (because code snippets need to be under CC BY-SA license, and GPL is not compatible with this). Which somehow makes the GPL not-as-free-or-open-as-some-might-think – at least in my book.

But hopefully all those discussions are now a thing of the past with the Apache License, so all that is left is the exhaustive debate if it wouldn’t be better to use that other “Open” license…EUPL, 2-clause BSD, 3-clause BSD, 1-clause BSD, MIT, EPL, LPGL, CDDL, MPL, or even my all-time-favourite, “The Unlicense”.

So, if you really believe Castle’s licensing was more ‘open’ than GPL I can only suggest that perhaps you should learn a little more about licensing

Been there, done that. If you really think that you understand fully what the GPL allows and disallows, you are at least one step further than the FSF.

Dragging this back on topic, one of the nice things about the relicensing is that it should make redistributing projects that build on parts of RISC OS a lot easier.

And indeed this is one of the reasons why the Apache License was a very good thing to choose – it makes “combining the work” with work under many other licenses very easy and unproblematic.

Spotted: https://bit-tech.net/news/tech/software/risc-os-re-released-under-an-open-source-licence/1/

https://www.theinquirer.net/inquirer/news/3064969/acorn-computers-risc-os-finally-goes-open-source

Very wrong in the details – read the last two paragraphs, which confuse the whole issue completely – but that is journalism nowadays for you…

“Bovine excrement” – “keep drinking the Stallman Kool-Aid” – sigh; Well I will say this, an Apache License means the project can be forked and no one has to care about such opinions any more or abide by the rules of a few individuals stuck in the 90s. Happy days.

The Register’s article has now appeared in Apple’s News app on the iPad and on the Mojave iMac under Computer Hardware.

Found another one in Deutsch :
https://www.pro-linux.de/news/1/26425/risc-os-unter-freie-lizenz-gestellt.html

Exciting times.

There are lots of possibilities and I’m looking forward to the future as RISC OS continues to evolve. That’s my focus today.

Personally I keep wondering about transcoding from ARM source code assembly language to C. But then again I’ve always found transcoders interesting….

If more of RISC OS were in C – where it did not hurt performance – then this would improve code understandibility & development. https://en.wikipedia.org/wiki/Decompiler. I’ve discussed this before. I need to make time to have a go at this really. I’ve seen it done for x86 code in a commercial setting more than 10 years ago now.

an Apache License means the project can be forked and no one has to care about such opinions any more

Yup. You can even GPL your copy (the FSF has a handy document telling you how to make lots of different things join the GPL collective), so, dream come true for you, right?
The bit in bold is, obviously, because that’s what everybody is hoping doesn’t happen. We need one main copy, not people having sulks and forking (OpenOffice comes to mind here).

a few individuals stuck in the 90s

Yeah, I believe that’s more or less how you dismissed me last time. Whatever. You keep telling yourself that your preferred licence is truly “free”. Whatever. I’ve done this argument already, I’m bored of it…

I’ve seen it done for x86 code in a commercial setting more than 10 years ago now.

[Moved to Aldershot]

Guys, could you take the discussions about GPL to somewhere where that’s relevant? It’d be great if we could keep this thread on topic…