Re: The Case Against Linux

[Jeff Licquia <jeff@luci.org>]

I noticed your article from a link from Slashdot (http://slashdot.org), 
an online news service with an admitted bias towards Linux.  As the 
co-founder of a Linux user's group, I must admit to the same bias; you 
should not be surprised that I disagree with you.  However, I thought I 
would comment on several of your points.

Your comments, by the way, are marked in classic Usenet reply style, with 
">".  For the sake of our group's discussions (which you can follow, if 
you're interested, from http://www.luci.org), the original article is at 
http://www.zdnet.com/zdnn/content/pcwk/1521/317585.html, and there is an 
ongoing commentary on Slashdot as well.

> But the use of Linux on so many Web sites merely reflects the nascent 
> frontier environment of the Internet. In the days before the Web became 
> ubiquitous, IS professionals saw its value when financial people still 
> didn't get it. The IS professionals, determined to get themselves and 
> their company on the Web, went for Linux, which was free and ran on 
> standard-fare Intel boxes.

This is very true.  Linux has often gotten in the door for this very 
reason.  I myself have introduced Linux in various places using this very 
strategy.

What you miss is the reason why it stays.  Often, I have seen people get
started on a Linux box, and then move to Windows NT once they get
"serious".  The prevailing thought is that, while this hacker OS is nice
and all, NT is the OS of choice for a serious Web site.  What surprises
them at this point is the *reduction* of performance and reliability they
see once their IIS site is up and running, especially considering the
vastly superior hardware they are most likely now using.  Some (like my
current employer, unfortunately) stay with NT and keep it propped up as
best they can, for whatever reasons.  Many more scrap it and return to
Linux. 

(BTW: While NT is a popular whipping boy, I have also heard similar 
stories involving SCO and NetWare.)

I will also point out that there are many other situations (besides Web 
sites) which bring this up.  Most of these have one difference, however: 
The Linux box goes in not to support something that finance considers a 
boondoggle, but to provide a needed service that some other product can't 
deliver.  In these cases (again, not to beat a dead horse, but most 
likely involving Windows NT), they turn to Linux as a "last resort" of 
sorts, often simply because there's a box handy and they're in a crisis.  
Often, these people are surprised by the smoothness of the transition and 
better performance.

As an example, we were having trouble with Windows NT RAS supporting our 
dialins.  The final straw came up when our president was unable to dial 
in and get critical information he needed while negotiating with a 
customer.  After replacing the NT box with Linux, we couldn't tell our 
IS director that the change had been made for three weeks (due to his 
schedule).  He never noticed (except, of course, that it worked much 
better), and even congratulated us for getting NT RAS to work properly!  
Now, that Linux box is totally insulated from the politics involved in 
our transition to an all-NT shop for one simple reason: it works too well 
to be replaced.

> But Linux is a communist operating system in a capitalist society.

This comment deserves some examination by itself.

Linux is not "communist" in the classic Marxist/Leninist sense of "from 
each according to his ability, to each according to his need".  Further, 
none of us believe that it's anti-capitalist (which Marxism and Leninism 
certainly are).  While it does have some communal attributes, its most 
obvious political parallel is libertarianism.  In fact, the Linux model 
relies more on the "invisible hand" of the market than does even the 
American economy, making it more capitalist, not less.

> Its 
> popularity is going to lead toward its fragmentation. There are several 
> for-sale versions of Linux available. For example, both Corel and 
> Interbase said they'd support Red Hat Linux. Does this leave Caldera in 
> the dust? What about the other for-sale Linux distributors, each of which 
> has access to the original source code, and each of which can modify it 
> to their liking?

To answer your question: no.  Except for Interbase (who licenses 
exclusively for Red Hat Linux 4.2 - it's not even legal to run under 
5.0), all software packaged for Red Hat will install and run under most 
other distributions.

There are two reasons for this: the Linux Filesystem Hierarchy Standard, 
or FHS, and RPM.  Because of the FHS, almost all Linux distributions are 
arranged the same way, so commercial packages fit right in.  And RPM, the 
package technology used by Red Hat Linux, is now a de facto standard of 
sorts, meaning that an RPM for Red Hat will install the same way on 
Caldera and SuSE.

The fragmentation argument is one that's been predicted by many, but the 
reality is that it hasn't happened yet.  The reasons are too varied to go 
through here, but suffice it to say that fragmentation typically only 
happens for technical reasons.  People who try to fragment for other 
reasons are rejected pretty quickly by the users, and the new project 
usually dies off for lack of interest.  And while the distro makers can 
fiddle with the source code, they're required by the license to 
distribute source for the changes, so proprietary extensions are nearly 
impossible given some legal vigilance.

> Linux may also run into problems regarding service and support. When 
> companies start merging their display-only Web sites with e-commerce 
> transactions, they're going to want high-end service and support. 
> Although media and Microsoft critic Joe Barr may be right that support 
> from the Linux community could be better than what Microsoft offers, the 
> community still does not have what corporations need.

Could you explain, sometime (maybe in a column) what it is that Linux 
needs by way of support?

Many Linuxers see this as a bit disingenuous.  If Linux support via the 
Internet is better than typical phone support (which I think we both 
agree is true), and if you can buy Linux support contracts from people 
like Cygnus for those high-end jobs, what else do you want?

What's worse is that the Linux community tends to see disfavorable 
articles (like yours) as challenges.  Witness the Linux desktop projects 
(www.kde.org and www.gnome.org), responding to the oft-quoted criticism 
that Linux is too hard to use.  But if there aren't any details supplied 
as to what's lacking (just "it's lacking"), then there isn't much we can 
do to fix it.

> The big problem with Linux is that it has no apparent direction. It's 
> in the right place at the right time, but its 15 minutes are nearly up. 
> No one knows for sure whether Linux will support clustering out of the 
> box, even though companies are building it themselves.

First data point: Extreme Linux, from Red Hat.  Basically, it's the
Beowulf clustering project (you know, the one that won the recent
prestigious supercomputing award for best performance in the industry)
packaged as an out-of-the-box cluster OS.  Price: around $30.  Considering
that packages like Wolfpack aren't even out of beta yet... 

Second data point: Fermi National Accelerator, in Chicago, has announced 
that they're building a thousands-of-nodes cluster to control the 
accelerator and process their data.  Guess what OS they're using?  Hint: 
it's not NT.

Linux has, so far, had pretty good direction from Linus.  Its success at
this stage has been largely the result of his leading the pack, from
spearheading the cross-platform movement to moving to ELF to advocating
threading to insisting on a desktop.  And, to reflect the pressures of
leading an increasingly popular OS along, he has begun to delegate, from
letting go of maintenance of the current stable kernel to allowing the
distros to lead on the libc5/glibc conversion to letting go almost 
completely of the networking code.  And he has outlined his vision for 
the next beta cycle of Linux, and it includes exactly one data point: 
kernel clustering support.  Stay tuned. :-)

Finally, if I may comment on one more point: Your whole article assumes
one thing that appears to be lacking at this point: an alternative.  Linux
arose precisely because of the problems apparent in the current commercial
OS offerings.  None of these appear to be changing.  Microsoft is still
more worried about its investors than its customers, the UNIX vendors are
acting like they're losing, and Apple is still floundering about in a
desperate attempt to stay afloat.  Novell, IBM, and the other server 
makers are in similar positions.  Linux's faltering implies that 
something else will take its place; since its current success is largely 
based on its replacing all the current players, what do you think will 
change?

Thanks for the opportunity to comment.  Feel free to respond or otherwise 
dialog.



--
To unsubscribe, send email to majordomo@luci.org with
"unsubscribe luci-discuss" in the body.