__/ [ Ian Hilliard ] on Friday 17 November 2006 20:50 \__
> Daveman750 wrote:
>
>> Operating systems, due to their nature, benefit from what economists
>> call network effects. This means that the value of an operating system
>> to an individual user depends on how many other people are using that
>> operating system.
This is only true if you count and consider the factor which is vendor
support for the operating system. The number of people who are involved in
developing and maintaining a platform can be thousands. You don't need
hundreds of millions in your workforce and feedback loop (testing). Look at
BSD, for example. Even RISC OS is still alive and many people use it for
most of their needs. It was even released as shared source recently. See
drobe.co.uk.
>> The incompatibilities between different distributions
>> of Linux lead to each distribution being, for some practical purposes,
>> a different operating system. While there is room in the market for
>> two operating systems, there is not room for the thousands effectively
>> created by this fragmentation of Linux.
There is no fragmentation. Linux is made out of packages that are shared by
distributors and system builders. Different packages suit different needs,
so herein you have a perfect system for the task at hand. No 'one size fits
all' approach anymore...
The fragmentation you speak about /might/ be choice, e.g. GNOME versus KDE,
lilo versus grub. They don't lead to forking. Moreover, their source is
public (open) so one can benefit from the other at any time. It saves the
user from that evil which is endless customisation (and yet it permits
everything to be changes, so it's not strictly an 'out of the box'
experience).
>> In fact, although I do not
>> approve of some of the more predatory tactics of Microsoft, one of the
>> silver linings in their domination of the operating system market has
>> been the ability of the industry to focus on solving problems in the
>> way that works best for Windows, instead of having to focus on the
>> general case. Below is a list of concrete problems that are caused by
>> Linux's fragmentation:
>>
>> 1. The difficulty of creating easy-to-install packages. A package
>> maintainer is left to either distribute the package as source and force
>> the user to deal with build environments, slow build times, etc. or to
>> distribute the package in a huge number of different forms, one for
>> each version of each distro.
>>
>> 2. Since every distro has a slightly different kernel, both in terms
>> of version and patches, it is difficult to impossible to create
>> easy-to-install, pre-compiled drivers for Linux.
>>
>> 3. Since the GUI in every distro has a different look and feel, it is
>> very difficult to give widely applicable instructions to a Linux user
>> on how to do things the GUI way. Despite the fact that Linux has had a
>> GUI for years, its usefulness has been limited by this. Instructions
>> tend to be given in CLI format, as this is the lowest common
>> denominator between distros. Also, given the variation in the GUI
>> between distros, application software cannot tightly integrate into the
>> GUI. For example, it is difficult to have an application installer
>> automatically put a link in one's Gnome or KDE menu each time a new app
>> is installed, at least without distro-specific packaging.
>>
>> 4. Important system files tend to not be in the same place on all
>> distros, making cross-distro scripting, such as creating install
>> scripts, relatively difficult.
>>
>> 5. Because of the above, when a person learns a given distro of Linux,
>> these skills do not generally transfer well to other distros, leading
>> to user frustration.
>>
>> In order to solve these problems and to make Linux a viable home/office
>> desktop solution, all major PC, peripheral hardware and software
>> manufacturers need to standardize on a single distribution. PC
>> manufacturers would put this distro on every mainstream home/office
>> desktop Linux system. Hardware manufacturers would make sure drivers
>> are easy to install on this distribution. Software manufacturers would
>> make sure that all software is packaged to install with minimal
>> headaches on this distro. Such a distribution would have the following
>> features:
>>
>> 1. A focus on stability (the doesn't break packages and drivers kind)
>> at the expense of being "cutting edge."
>>
>> 2. A long time between releases, such that kernels and libraries
>> aren't changed until the difference between versions becomes
>> substantial, and thus things break only once every few years instead of
>> every few months.
>>
>> 3. Standardization on a single default GUI desktop environment and
>> tightly integrating it into the operating system.
>>
>> 4. Given the long release time, possible point releases that update
>> only application software and not kernels, libraries or anything else
>> that might break anything.
>>
>> While this would limit choice in the Linux field, this is a necessary
>> evil, as it would allow for the greater good of giving users a choice
>> that's neither proprietary (like Microsoft) nor too fragmented to
>> benefit significantly from network effects.
>
> Nice, well laid out article, based on misunderstanding the Linux ecosystem.
>
> 1) The vast majority of Linux distros are simply to prove a concept. These
> distros get at best a niche following. The effect of these distros is to
> show what is possible and where the ideas are good, they are adopted by
> others. This is bringing the excitement, that was DOS to a whole new level.
>
> 2) The majority of distros can trace their pedigree back to one of the big
> four: Redhat, Suse, Mandriva & Debian.
>
> 3) A lot of programs, such as Firefox et. al. will run on pretty well any
> distro.
>
> 4) The major distros have repositories that provide most of the software
> that is required by most people.
>
> 5) As the majority of code written for Linux is open source, it generally
> only requires recompiling to get to run on a specific distro. This code can
> then be distributed in binary format.
>
> 6) Non open source code is generally written for a specific platform and
> that is the platform on which it is tested and is to be used.
>
> 7) There is enough similarity between all the distros for Linux to be a
> generic term for a Unix clone, based on the Linux kernel, that provides
> stability, reliability and top value for money.
>
> It is up to each person to decide which distro best suits their needs and
> they can stick with that. But, unlike with things like Windows, should you
> no longer wish to use that distribution there are many others, which are
> probably just as good if not better for the users specific needs.
It's rather amazing how living a certain single approach can lead many to
blindness. I guess that many can assimilate a piece of software to something
rigid like a fridge.
|
|
