[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: CNET News Article

From: "Don Maxwell" <don maxwell usa net>

> This morning's lead article in CNET News.com entitled "Red Hat overhauls
> flagship Linux" had references to Fedora which I found to be be
> disappointing.  They quoted from users in an email list that I believe
> may have been confused about Fedora.  Perhaps some folks on this list
> can clarify a couple of points.
> First, there were concerns noted that Fedora will be a fast-changing
> distro, not conducive to a stable working environment.  Granted,
> wholesale upgrades from one release to another might break things and
> could consume resources to accomplish the task.  However, I am unclear
> as to whether the commenter's were concerned about ongoing upgrades
> within a release or were concerned about the need to upgrade all servers
> and desktops perhaps three times a year.
> What should a user's expectation be with respect to planned upgrades and
> releases?  

With the level of difficulty for providing a really bug free collection
of software for something such as a Red Hat distribution (or a Windows
distribution, which is actually smaller, for that matter) is pretty much
"not going to happen" even for one specific CPU and stepping and one
specific motherboard and collection of cards. Then when you consider the
myriad of hardware problems in the hardware this software is expected
to run on the only sure thing is that "bugs happem"

Given that bugs happen figure that a percentage of these bugs, perhaps
a high percentage, will represent security holes. (BSD distributions
that are built for security first still experience security problems
from the critically needed add on tools such as browsers, servers,
and user applications. It may be that the basic BSD structure minimizes
this. But it does not eliminate it.) Security patches can keep up with
the holes for awhile. However, as Microsoft has found out many security
patches themselves cause system instability and further holes. At some
time the probability of introducing problems becomes too close to the
probability of eliminating them. At that time the distribution needs to
face a clean and humane death. (Sadly Microsoft distributions seem to
come off the production line in that state for many system uses. But
that is an extreme example. Even Linux gets there eventually.)

Linux gets there because it gets boring to maintain the same code day
in and day out. So the maintainers redesign it. (2.0, 2.2, 2.5, and
now 2.6 kernels enter the picture. New incompatible compilers and
libraries appear, and so forth.) At sometime the new hardware people
wish to use is too incompatible with the kernel to run it for anything
"user-ish" and probably even for server uses in extreme cases. So the
whole smush needs to be updated again with a whole new distribution
and we restart the bug cycle with a new range of bugs as well as many
more that were there in the old code but just not found yet. Sometimes
new technology for security is developed that requires a whole new
means of compiling the application code as well as system code. Whatever,
eventually old software is not going to be maintained. I rather expect
that the life of a Microsoft OS with patches may exceed the life of a
Linux distribution due to human factors. At Microsoft the people are
getting paid to work on the old stuff.  Money talks, people listen,
and those people whore around in old code for the income it means.
Linux offers little or no income to most of the developers for the
various tools and programs that are delivered with a distribution.
Linux offers little or no income for most kernel developers who might be
moved to maintain proper backwards hardware compatibility. So it does
not get done. Linux pays most developers in 'ego points'. The best ego
points come from developing for the neat new stuff. And even there
money talks. MS can manage to get some of the niftiest newest "stuff"
so they can maintain compatibility with hardware as it comes out or
even well before it comes out. Linux provides a disencentive for companies
like Nvidia to develop Linux drivers. They have, they feel, proprietary
money earning technology they have no intent to make available to their
competitors. So they must release in binary. And the result is made a
pariah in the Linux world. It's somewhat amazing they even bother. Linux
is definately low on their priority lists. So Linux keeps up with new
hardware as it comes out and developers can get their hands on the cards
and documentation for the cards. It will lag. And when that server needs
upgrade old kernels and software libraries may not handle the new machine.
So you must ungrade. Three years seems to be close to the sustainable
maximum life for any serious distribution mostly on a human factors level
for Linux.

I wonder how this affects the <choke> "Total Cost of Ownership". I do
note that successful hacks are a major often neglected item in the TCO
equation. It is also a very positive side of Linux as compared to the
results of the paid (and pained) minions of the Gates fortune.

> Second, one comment described Fedora as "possibly full of breakage."   I
> think the poster may be theorizing about potential future upgrades.  I
> certainly hope that the comment was not made on the basis of the
> stability of test releases!

See the above discussion. And note that "full breakage" may simply mean
a recompile of the old code in most cases. {^_-} Journalists tend to be
alarmist by profession. It's their ego boost item. It also pads their
bottom lines.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]