2.3. Linux and other operating systems

2.3.1. Linux and other proprietary Unix (HP-UX, Tru64, AIX, Solaris, Irix)

The situation taken by all the hardware manufacturers places Linux today in the entry level (when it's taken in account) and their own Unix system in the middle and high level. Reality is sometimes more cruel than the situation wanted by the marketing department :-). Technically, a Linux distribution has no cause to be envious of the manufacturers Unix solutions, except for the moment the set of commercial applications and some high-end hardware configuration support. And we can often see that users dope their Unix workstations with free software to complete their usage.

To be precise in the talk, we should consider manufacturers Unix solutions as split in stations and servers.

On the station side, there's no doubt in my mind: for a customer, the only reason today to buy one is linked to the availability of a software or a hardware which wouldn't exist in the free environment, or due to intrinsic performances of the machine. In the first case, we can see that this argument should have a short life time, because logically every software editor (except maybe hardware manufacturers) and every hardware manufacturer has interest in porting their applications or allowing the use of their hardware on all the major environments in the market. As shown by Oracle, Informix, Sybase, ... Linux becomes today one of the major environments. In the second case, the difference in term of performances between the Linux dominant platform (IA-32) and the other competitors decreases and will disappear with the generalisation of the IA-64 architecture, which seems to be adopted largely by a majority of actors. More over, Linux is often available natively on the processors of these actors (Sparc, Mips, PowerPC, 68xxx, Alpha, Crusoe, PA-Risc). I think Linux may represent the famous unique and standard Unix that everybody dreamed of from a long time, without succeeding to impose it (the open and free characteristics brought by Linux are not without influence on that). I think that at the end hardware manufacturer Unix workstations will be restricted to some niches such as high end computation, virtual reality, ... and that as long as solutions are not available on a free environment. Linux offers all the functional qualities of the other Unix systems, and thus of the workstation, on a more various hardware platform and potentially at a better price, if on PCs. So it is the natural choice of every computer engineer with a strong Unix culture (old customer of the workstation) who will prefer that solution to the migration to a Microsoft system typically. With IA-64, we can see that Linux was the first operating system available (including graphical support).

On the server side, in addtion to the points mentionned previously for the stations, problems are more complex. Ram, disks, processors capacities, extensions of every kind make them difficultly repleacable by machines with a IA-32 architecture for example. More over, certain solutions such as high availability clusters for example have just been put in production in a Linux environment, compared to years of operation under Unix. The other brake is often linked to investments already done around software solutions deployed on these servers. Whatever their natural life time is much higher than those of the stations. Changes will thus be made more slowly in that domain. Here we can consider rightly Linux solutions as an entry/middle level solutions, when hardware manufacturer Unix servers are the middle/high level. The introduction of Linux in place of these machines will begin only with a massive availability of applications, mainly in the management sector. Again IA-64, with increased SMP capabilities (16 processors today), higher bandwidth, ... will make things change.

The advantages of the hardware manufacturer solutions, explaining why they are so often chosen when applications are critical, are linked to the homogeneity of the solution (hardware and software mastered by the same entity, which can't invoke a third party in case of problem), and to the support and maintenance guarantees furnished.

Finally, there is not so much antagonism between these systems, because they are full cousins. Their association allows today to computer teams "pro-Unix" to have solutions from start to finish, without having to lose in functions, as it's so often the case with other operating systems available for personal computers.

2.3.2. Linux and SCO/Caldera

The comparaison between Linux and the SCO systems seems to me quite unbalanced. First, all the previous points are vaild here also. More over, SCO isn't a hardware manufacturer, so the homogeneity advantage disappears. The IA-32 Intel platform is supported by both systems, so cost is identical. On the other hand, the software solution has a disproportionate cost (few software provided in the base install, thus a lot of expenses to extend, as well as to increase the number of users). Having to manage one OpenServer, I can say that performances are far beyond those of a Linux system. More over, its conception is older, abounds of symbolic links which makes management complex. The hardware supported by SCO is less numerous than those Linux supports. Only stay as an advantage the installed base and the set of applications available. But for how long ? Caldera now owns both SCO software (Openserver and Unixware) and its own Linux distribution. But instead of trying to free technologies from SCO, they are proprietarizing their linux (1 license per server).

2.3.3. Linux and Windows NT/2000/XP

The comparaison is here more difficult, because Windows NT/2000/XP isn't an open system, as the precedings, which is already redhibitory for certain users. An excellent comparaison was made by John Kirch between Unix and Windows NT Server. I recommend to people searching to have an enlightened opinion on this subject to read it; it's updated regularly and was written by a specialist of both Microsoft and Unix operating systems. Financially, obvious advantage for the free software. And it's more obvious, as for SCO, if you consider the set of complementary software you need to use a server. The author evaluates the difference from 1 to 100 all the same ! Technically, either on the functions provided or on the reliability, the administration, the performances, the hardware supported, and more over the security, Unix systems and particularly free systems outperform what is proposed by Windows NT/2000/XP. The fact to have a GUI non independant from the kernel contributes greatly to the instability of NT/2000/XP, because it's more difficult to avoid errors in a GUI (there is no mastering possible of the user comportment in front of it) rather than in a kernel.

Which are the real advantages of Windows NT ?: the marketing power of Microsoft which persuades the world that computers equal Windows and which leans on the enormous installed base; its office applications (a monopolistic situation on the market) only available in this environment; the confusion maintained between the various flavours of Windows (95/98, NT, 2000), and between the server and client functions; its agreements with the biggest hardware manufacturers which often require them to provide a Microsoft system with their platforms; its technological initiatives to occupy the market in first, based on proprietary code and that often without respect of known or documented standards, or its policy to embrace for killing a movement (described in the "Haloween documents").

The lack of hegemony in the servers sector is the best reason to hope that a plurality of solutions may exist in the future for computers users, also at the desktop. The new XP licensing schema, the passport tool, the registering mechanism during installation are real brakes for users thinking to migrate from a current platform to a new one, as well as the hardware costs linked. In that perspective, an opensource solution should and will be more and more considered.