Amigactive Magazine Dissects Amiga's Linux Announcement

11 Jul 1999
By Andrew Korn.

Amiga's announcement (July 9th) that the next-generation Amiga OS would use a Linux core has thrown the Amiga community into chaos and confusion. Just a day after the announcement from QNX Software Systems that their updated Neutrino / Photon multimedia desktop OS originally planned for the next generation Amiga was nearing completion, this sudden change has seen the positive momentum of recent months hit the buffers.

The conclusion many commentators have jumped to is that Amiga have been drawn to the flame of the Linux media hype, and moved to an inferior, less revolutionary technology. However, upcoming revelations in the promised technology briefing and at the World of Amiga and Amiwest Amiga shows are likely to make the reasons a little less obscure.

The long-speculated rift between Amiga and QNX looks like it may be turning into yet another kernel war. QNX Software Systems seem keen to go ahead with their domesticated Neutrino OS with or without Amiga. They have offered developer kits to Amiga developers interested in working on their system, and offer all the neat features we have all been so intrigued by in the eight months since the initial partnership announcement. Amiga, holding their cards close to their chests, have yet to explain why they have moved away from this.

Amiga's announcement of the Linux partnership was clearly prompted by the actions of QNX on the previous day. This information was originally meant to be presented as part of a 5-7 page technology brief explaining much of the reasoning. Out of context, the announcement has caused a certain amount of panic in the Amiga community.

There are several important points that have not been fully understood in this situation. The first is what, exactly, this means. Amiga are not simply planning on producing a Linux box with their own GUI, running on x86 - they are planning on producing a unique OS which sits on top of the Linux kernel and utilises some revolutionary hardware expected to be announced on the 24th of this month. Amiga assure us that the AmigaSoft Operating Environment will not be just another Linux distribution.

Amiga will presumably be adding a lot to the Linux kernel - certainly a GUI system, judging by the emphasis on human interfaces engineers in the jobs list published on the Amiga website. It is probably safe to assume that any decent human interface talent should ensure a friendly filesystem, and so on. Other talents strongly emphasised in Amiga's requirements are networking, electronic transactions, and multimedia technologies. We can assume that these indicate further significant developments on top of the kernel. The "AmigaObjects" system, which allows network devices to share some processing demands and the increasing stack requirements necessary for multiple high bandwidth streaming applications may also imply development of the Linux TCP/IP code, which is not well regarded as it stands.

The Pros and Cons of Linuxing

There is no doubt that at the moment the Neutrino kernel from QNX Software Systems appears more at home in forward looking arenas such as realtime multimedia services and pervasive networking, but we have to remember that we are not necessarily comparing any more than the glue that ties the operating environment to the hardware. At the moment, we simply do not know the reasons why they changed the glue. In general the innovative features that have been publically emphasised for the NG Amiga may appear to be QNX innovations, but the simple truth is that as yet we do not have enough information about the innovations that Amiga are bringing to the table to know which are the most important.

The move to Linux is a difficult one to measure at the moment, because while we have a good idea of the innovative aspects of Neutrino, there is as yet a lot to be revealed as to where Amiga are going with Linux. One thing that seems clear is that we should not assume that all the features Amiga have been discussing over the past few months will have disappeared - far from it. Some people have taken the latest announcement to be yet another change in direction from the company, yet it seems clear from the announcement by Amiga president Jim Collas that this was being considered ever since he joined the company in February this year. As such it can be considered just another piece of the same change in direction we have been hearing about in the last few months.

There are obvious marketing advantages in Linux, notably that we can now be pretty safe in assuming that the new Amigas will run software like Quake 3 and Corel Office from the outset, and will benefit from large scale and well optimised driver development. No doubt the Amiga executives will have considered the publicity advantages that might accrue from the current media love affair with Linux. They will also have considered the advantages to be gained from being able to tap into the Linux community as well as the Amiga community - a partnership of OS communities that seems quite appropriate. The credentials of the new Amiga platform for shareware will be simply unsurpassed with the meeting of Amiga developers and Linux resources, but it will be interesting to see how the commercial requirements of Amiga will tie in with the open source traditions and philosophies of Linux. Amigactive Magazine asked Linus Torvalds, author of Linux, what he thought of the issue:

Superchip Revealed?

The picture forming of Amiga's plans is of a cheap, powerful multimedia computer, compatible with important upcoming technologies such as DSL, Universal Plug and Play, Havi, Java and Jini. It will have a Linux core, its own distinctive user interface system and some exciting pervasive network computing technology through the AmigaObject system. Some of these developments will presumably be fed back into the Linux open source through GPL, while other elements will be unique to the Amiga. Will it be as good as it would have been with the Neutrino kernel? That depends on what Linux brings to the table. The reason Jim Collas gave for the Linux switch was that he did not believe that going it alone with QNX was going to enable them to keep up with the rapid advancement of technology - this might be assumed to be a reference to the driver issue, but I suspect there is more to it than that.

There is a very interesting line in Jim Collas' "Response to Linux Feedback" posting on the Amiga website:

What does this mean? "Better than any other computer in the world" is quite a claim - and also quite a clue.

Jim Collas recently listed several of the chips in the running for the Amiga CPU. One of the chips mentioned is a project from a well funded and well staffed Santa Clara technology startup, Transmeta. While their project has been hidden behind a masterfully engineered veil of secrecy, what little is known or widely speculated about Transmeta makes it fit very well into the Amiga scenario, propelling it clear of the pack in the race for most widely rumoured Amiga NG hardware.

Transmeta is listed on the web site of investor and original Microsoft co-founder Paul Allen as a company whose core business is "Alternative VLSI engines for multimedia PCs." Their sole patent - number 5,832,205 - goes under the obscure title "Memory Controller for a Microprocessor for Detecting a Failure of Speculation on the Physical Nature of a Component Being Addressed" by Edmund Kelly, Robert Cmelik and Malcom John Wing.  This patent describes a 'code morphing' technology that apparently converts CISC, RISC or even Java instructions to VLIW processing on the fly. The clues are slim, but common guesses suggest a relatively cheap VLIW processor which runs alien code at speeds comparable to a native processor. In practise it is a fair assumption that a VLIW architecture would deliver a significantly better performance with code acting on a data stream (such as multimedia operations) while integer performance would not be so optimised.

Now, although all of this is just speculation, tying the threads together we can assume that the multimedia PCs intended for this "alternative VLSI engine" will provide roughly similar integer performance to other processors with a significantly improved multimedia performance, along with effectively native JAVA code operations, and - given the presence of Linus Torvalds on the staff of Transmeta - quite likely a low level support of Linux as well. Further hints suggest that the processor will require a lower transistor count than other architectures, allowing smaller dies than comparable technologies (one quarter the size of Intel processors has been mentioned) and low manufacturing costs. Sound familiar? A partnership between Transmeta and Amiga would certainly set the computer market alight with excitement - but, of course, neither company is saying anything right now.

Two Kernels are Better than One?

So, where does this leave QNX? At the moment that is hard to say. They have an impressive, brand new operating system with nowhere - at the moment - to go (currently it only runs on standard x86 architecture). It seems strange, given that both Linux and QNX are POSIX systems and much has been made of Amigas plans for a high level of portability, for Amiga to drop QNX at this stage. Until Amiga clarify the Linux situation, we'll be left speculating. In the meantime, we may suddenly have two new operating systems to consider. The pessimist may see this as the cause of an inevitable split in the market, whilst the optimist will consider it a welcome safety valve should San Diego fall into the sea, taking Amiga's Headquarters with it.

It isn't impossible to envision a future where Amiga and QNX scrap for the pieces of the Amiga community, but it would be foolish for the community to self-destruct at this stage. It is also easy to see a future in which an Amiga layer can run on either Linux or Neutrino systems, each offering a different set of advantages. Trying to second-guess the future without knowing the reasons why Amiga made the change is ultimately futile. Amiga may have driven down a road which looks like it runs through a much less scenic part of town, but it is the destination we should be judging, not the route. Certainly this route should help Amiga bypass many of the problems of kick-starting an entirely new OS.

When Amigactive Magazine attended the Amiga conference at Heathrow a couple of months ago, we were presented with an intriguing - if incomplete - vision of the future. That vision has not changed - it is still just as intriguing, and it is still just as incomplete. We know that Linux will be the OS kernel, but it won't necessarily be the current incarnation of that kernel, so we don't even know the full kernel functionality for sure. It may not be as compact as the Neutrino kernel, but Amiga have promised us a "clean and efficient Linux configuration as a standard part of our operating environment."

The main worry is that Amiga may have entered a market they can't lead. Other companies have decided that Linux is the way forward, and are planning user-friendly Linux based systems. For Amiga to succeed they will have to ensure two things - that they are more innovative than the alternatives, and that they stand out from the crowd not just in terms of power but also in providing a viable commercial market. Many Amiga developers will be worried that they are going to have to compete with huge free software developments, but we are assured that Linux itself is only a subset of the Amiga functionality. The real innovative power of the next generation Amiga will need to be addressed directly, so Amiga software running on an Amiga system would offer significantly more power and functionality than pure Linux software.

So, can Amiga produce a revolutionary Linux machine? Linus Torvalds told us his feelings on the announcement:

Of course, asking Linus whether Transmeta would fit in there too would be a futile exercise - he's far too discreet to say. However, it seems clear that any company who can combine Linux with a user interface as friendly as the Amiga's, the promised innovative developments that have survived from Amiga's announcements of the last few months, and a truly impressive, low cost hardware platform to run it on is going to be in a killer position for the future. At worst, Amiga have traded some part of the technical improvements for a more potent assault on the hearts and minds of the computing market. At best, they may be aligning themselves with the means to sweep the computing industry in its wake.