Community: Rebuttal of "RedNova: WinServer Vs. Linux Study Generates Fire"
Jun 14, 2005, 23:30 (19 Talkback[s])
(Other stories by Peter Surda)
Full Text Search: The Key to Better Natural Language Queries for NoSQL in Node.js
[ Thanks to Peter
Surda for this article. ]
I read this
paper on Sunday before it was posted on LT. It says, among
other things, "we are genuinely committed to an open and
professional dialogue about technology." So I wrote a rebuttal and
sent it to the authors via email. I thought perhaps making it
public would make some sense too, there is nothing personal in it.
So here is it, straight out of my outbox:
I just read your paper "Evaluation No. 1: Enterprise Operating
Systems" and immediately had the urge to reply and correct several
First of all, however, let me state that I am not "officially"
impartial. I have my own Linux distribution and provide commercial
Linux services. Furthermore, I only have limited experience with
Microsoft Windows. However, I will try to stay objective.
I would also like to say that not all mentioned Linux problems
were assessed wrongly. There are some very valid critique points.
However, I won't elaborate on this, neither on the areas I don't
have enough experience in to make an objective analysis (such as
LDAP). Also please keep in mind that I don't get paid for this, so
I can't spend much time crafting this response, and hence some
parts may not be completely clear or stylistically correct.
I would like to begin with the quotation:
"In addition to being a Microsoft Certified Systems
Engineer, our Linux tester was also certified an LPI by the Linux
Professional Institute. Most recently, he has spent the past five
years as a Linux consultant..."
Several passages of the paper shed some serious doubts about
this claim (for example, both graphical remote administration and
Automated Software Distribution & Maintenance are provided in
both mentioned Linux distributions and are easy to use, see below).
As I have no reasons to suspect the mistakes were intentional, the
only conclusion left is that the person evaluating Linux had in
fact very little practical experience with it.
Most of the problems were in my opinion caused by what I'd call
"handling Linux as if it were Windows," i.e., using the same
methodology, processes and infrastructure. Every experienced Linux
administrator will tell you that this leads to a failure. I have
seen this happen many times. This means that many of the perceived
mistakes were not Linux' fault, but fault of the organisation, who
thinks in Microsoft. In order to use Linux effectively, you have to
think in Linux. The same holds in the opposite direction too: if I
wanted to setup and operate a Windows network, I would almost
certainly fail because I would want to treat Windows as if it were
Linux, and I already realized several times that this simply
doesn't work, but only after I wasted many hours of my time.
For starters, in many situations of making a company-wide linux
desktop deployment, it makes much more sense to use thin-clients.
Both mentioned linux distributions provide tools to do so (X, VNC)
and there are additional products available (for example NoMachine
NX). Using thin clients instead of "normal" desktop leads to
dramatic cost reduction both in downtime and administration. I have
been using a thin client as my primary "computer" for about 2.5
years both for work and private use. The machine I usually use is a
real thin client (specialized hardware). This response was also
created "on" it. Additionaly, wherever in the world I am and
regardless of what OS is installed locally, if I want to work, I
just SSH into my server and reattach the VNC session over the SSH
tunnel. In Linux use vncviewer and openssh, in Windows also
vncviewer but putty. So I am confident that using thin clients
should work in almost all typical situations.
Furthermore, you correctly point out that setting up and
administering linux requires more experience. I find it however
strange that in light of this you didn't contact a professional. A
professional can set up typical configurations much faster.
Another aspect of "Linux thinking" is automation. Experienced
linux administrators automate typical and recurring tasks, so that
setting up new machines or administering larger numbers becomes
cheaper. Linux has all the necessary tools to do so. These
economies of scale only show up when you actually have such large
numbers and know how to automate, and are not reproducible on a
small scale. I know this from personal experience, I administer
several dozens of machines and co-administer many networks, the
largest one consisting of about 1400 computers.
In general you can make use of this by outsourcing certain parts
of your IT to a professional company that specializes in Linux and
has the tools for automation. This way you'll save money and don't
have to worry about understanding how LDAP or pptpd works.
Now to the specific points.
This seems to suffer from the same "Microsoft thinking" problem. It
is there, and it is in my experience far superior to anything else.
I often communicate directly with the developers and we work on
fixing problems on a low level, together, usually within hours of
noticing the problem. And it isn't anything rare or new, this works
with almost any open source project, and has been working for
years. Can you provide me an example where you found a bug in
Microsoft's products, and were able to talk about it directly with
the responsible people (and, god forbid, they actually fixed it)?
However, a lot of inexperienced linux users don't know how to
participate in this whole process properly, and hence don't reach
solutions. The are either flooded with information they can't sort
(as your paper mentions), or ask the wrong people, wrong places,
wrong questions. I have been already thinking about this and am
planning to create a document presenting this process to a beginner
in a way that helps them get the most of it. It would consist of
two parts: analysis/description, and a checklist for those who wish
to participate. This will allow to promote the community and clear
"Automated Software Distribution & Maintenance"
As I mentioned, both distributions provide tools for doing so. I
have little experience with SuSE, but the tool you're looking for
is yast. I also haven't used a recent Red Hat product, but Fedora
(which Red Hat is based on) provides no less than three tools: Red
Hat Network, yum and apt. I have experience with all of these, and
didn't notice any problems. Other distributions have these or
similar tools too, for example, Trustix uses "swup."
"Installing & Configuring Printers"
I must agree to some extent that linux printing is a mess. However,
there is an easy workaround for typical cases: set every client to
use a PostScript-compatible printer, and let the server's spooler
deal with the conversion. That way, the client doesn't even have to
know what kind of printer it is and doesn't need any drivers
(PostScript is the default). I tested this several times and it
As I said before, this whole area can be solved by using thin
Apparently you missed "webmin," which should do exactly what you
need and is web-based. I don't know if it's in the mentioned Linux
distributions, but installing it isn't difficult. I'd also like to
point out that an experienced linux administrator has other options
too, for example I personally don't use webmin.
While I confess that automatic configuration backups as Windows has
some merit, I personally don't have any problems with backing up
any of the Linux machines I administer (there are several dozen of
them, scattered over two countries). As for the "client backup,"
this isn't necessary on thin clients.
"Desktop Security Management"
Besides the repetitive "use thin clients," in this topic I sense
"Microsoft thinking" again. If you want to disable certain client
functions, then don't install the software, easy as pie (in Windows
you often don't have this option). A firewall on the client isn't
required, because a proper linux client machine doesn't run any
services (besides SSH, which should be restricted to
administrator's access upon install, and X, which should be setup
in a way that it isn't accessible from the network), and isn't
infected with spyware and worms. As for the automated updates, as I
already mentioned before, there are several alternatives and you
don't need any paid subscriptions for them. Activating them is
easy, in fact Fedora does this already so I assume Red Hat solves
this in a similar way.
I admit here, setting up pptp/pptpd isn't easy. But as I said
before, when used on large scale, you can automate both client and
server side and save time (my Linux distribution "Route Hat" for
example does this). Furthermore, there are other options that are
easier to maintain, for example cipe or openvpn. And the universal
tool SSH, which can be used to tunnel any TCP connections. Neither
of the alternatives needs root privileges for using (but as far as
I remember, cipe and openvpn require them to be set up initially,
but this can be automated upon install). Activating/deactivating
these VPN can be bound to a script on user's desktop, so that it is
a matter of clicking.
"Remote Server Management"
Repeat after me: there is graphical remote administration for
Linux, and has been around for many many years. X is by design
network aware, the application doesn't care where you're sitting.
More recent is VNC. Each has its advantages and disadvantages, I
personally use both. A security-aware administrator would tunnel
both over SSH or other VPN of course. But the bottom line is, it is
easy and typically works automatically ("ssh -X user@server
application-name"). And of course there is the already mentioned
Webmin. I'd like to point out however, that I personally extremely
seldom administer servers by the means of a graphical interface, so
whether it is actually required depends on the administrator.
"...it was difficult to delegate server administration to
The tool for this is called "sudo," and is present in basically all
Linux distributions. Setting it up may require some expertise, but
as I said before, think economies of scale. Furthermore, the
"webmin" I already mentioned twice should be able to do this
I omitted several parts, which was as I stated before either
because I agree with them or because I don't have enough experience
in that areas.
In summary, I consider your paper usable to some extent, but it
should be relabled "If you want to use Linux but still think in
Microsoft, this is how you will fail, so don't do it." In case you
would like to make another Linux evaluation in the future, I
recommend hiring a real Linux professional, not just someone who
has a certificate.
WinServer Vs. Linux Study Generates Fire(Jun 13, 2005)