Interview of Gratien D'haese by PANDOR for linuxzone.cz
Do you remember what your first touch with Linux was? When did you find
out that Linux exists, how did you start to use it? What was your breakpoint
for deciding to be a regular linux user?
I started to follow Linux news since the end of 1992 on some
Usenet newsgroups (I forgot which ones). If I remember well my first touch
was the Linux Nighthawk 0.99 release from March 1993, which was a tiny
linux on a single floppy (just booted it up to verify). In those times
one had to download linux as series of floppies (10 or more) to get a complete
distribution. My main interest to start with Linux was to study the internals
of the kernel. The kernel source was only 500 Kb big then! Secondly, I
was excited by the network capabilities and promoted Linux a lot within
the company I worked for. I even started to mirror Linux and GNU sources
from the Internet for internal use only.
I fell for Linux from the very beginning and decided to use it internally
the company as a dial-in server for remote system management. In those
days it was not allowed to use Linux in a professional environment, but
I hided the box under my desk and it worked for years without having any
problems!
Furthermore, I used Linux as a testbed for my papers I had to write
during my evening education (Informatics).
Used Emacs and Scheme day and night for weeks...
As for the linux starters and newbies. Do you think there is a sufficient
amount resouces and helpful advanced users to support them in the first
stages of linux exploration and usage?
Today Linux has become an excellent Operating System with splendid
graphical interfaces (KDE, Gnome). I cannot think of one tool that is not
available on Linux that is available on Windows. I agree sometimes it takes
some time to find something that suites your needs, but that makes Linux
so exciting. Furthermore, I always supported the Free Software movement
(and Open Source organisations).
To give an example, in 1994 the Belgian Unix systems Users Group (BUUG)
started with a Linux Special Interest Group (Linux-SIG) where I was one
of the key founders. The main purpose was to present Linux to the Belgian
professional users and the presentations were a big success. Remember,
in 1994 the Internet was only for the happy few under us and downloading
a Linux distribution took hours over 19200 baud modem!
Therefore, we made Linux and GNU sources available on CD-ROM for our
members. Today, people don't have to go to conferences or workshops anymore
to get excellent information on Linux (starter or expert information),
it is all on the Internet. A search engine brings it all to you (maybe
a bit too much).
Resources enough! The only problem I have with the Internet is that
I miss human contact as we had in the mid 90s during the very busy workshops.
We have noticed that Linux workshops are not fashionable anymore,
more and more Linux Internet portals have been setup, in my opinion
however this is not the same. People should meet to discuss and chat on
Linux and other matters.
During your tries with linux, you have surely tested a lot of distributions.
What are your feelings about them, for which one have you decided and why?
I have used so many distributions since 1994 (Slackware, Debian,
Redhat, Suse, Mandrake, and others) that it is difficult to say which are
the best. They all had pro and contras, but today I use mostly Redhat and
Mandrake on my production systems. On my test systems I have tested many,
many more (even UnitedLinux) as a testbed for mkCDrec.
I find the idea behind UnitedLinux very attractable where the best
of different Linux flavours are unified in one distribution, but internal
it still feels like SuSe. Companies like Redhat, Mandrake and SuSe can
only survive in the long term by making alliances with big cooperations,
like e.g. HP, IBM, Oracle, SAP to name a few, by adding more high-end server
features in their distributions. That is why I think (and this is my opinion
only) that closed vendor operating systems will disappear in the long term
as it is becoming too expensive to keep up with the development to add
new features. The Open Source model is much more suitable for these purposes.
Just have a look at IBM for example as they have plenty of software engineers
working on open source projects. They only do this as they belief that
it will pay back over the years.
You have worked on several Linux projects. Can you describe them? When
did you start the development and why?
My first Linux project started in 1994 and had as purpose to
build an complete site mirror on CD-ROM which was browsable via the Mosaic
browser (predecessor of Netscape). The Louvre museum was on it and I had
to build my own HTML syntax and dead link checkers. Thereafter, came the
Linux95 project, which was a BUUG tailored Linux distribution (I remember
it was released before Windows95!).
In 1997 I demonstrated a Linux server to act as boot server for Network
Computers including running Java applications such as Corel Office for
Java on one of our Linux-SIG events.
Did plenty of work around ISDN4Linux, firewalls, etc... See our web
pages at http://www.ota.be/linux/workshops/
The most popular of your projects seems to be mkcdrec with a very good
approach to storage backup and cloning. What were your objects for an internal
architecture and integrated features?
Mid 2000 I was building a Linux based firewall for a customer
and I needed a quick and dirty way to restore to system after a disaster
such as a break-in, broken disk, etc. I found no software on the Internet
that fulfilled my wishes, therefore, I started coding from scratch a disaster
recovery solution loosely based on the ideas behind Ignite/UX from HP-UX.
It was clear right from the beginning that I needed a bootable CD-ROM including
the backups. In principle the idea behind mkCDrec is simple: when booted
from the rescue CD one should get an environment which is capable to repair
a broken filesystem to name something, or to restore the complete system
from scratch in the worst case. Therefore, I needed some good tools, such
as BusyBox (swiss knife of integrated Unix tools) and home-brew scripts
to restore the system if needed. These scripts are partial build during
the making process of mkCDrec and collect lots of information of the system,
which is needed afterwards when one is booted from the mkCDrec CD. Typically,
everything one finds under /etc/recovery is made quite dynamically by mkCDrec.
It depends from system to system what you see beneath the /etc/recovery
directory tree.
What do you think about the available methods of backing up data in
a secure and reliable way?
mkCDrec uses GNU tar to backup your partitions, but has the
possibilities to encrypt the archives with openssl. One can also use ssh
to backup your data to a remote tape drive or site.
Surely, some problems occured during development. Can you describe
some
of the them with your clues?
As the project evolved over time I got more and more user requests
for new features. Typically, new code tends to break something else. During
the last months lots of problems occurred with BusyBox, e.g. it did not
reboot anymore, and the shutdown command was broken too. Therefore, I had
to write a small script to do that (alt-ctrl-del). Meanwhile, these problems
have been resolved by the BusyBox team, but now the Control-C is not functional
on the first console (on the other virtual TTY's Ctrl-C does work). In
the latest BusyBox-0.60.5 version I have trouble with the 2th, 3th virtual
consoles. Therefore, I need to dig into the code of BusyBox too to try
to find where if goes wrong and why. I also submit bug reports to the BusyBox
team to improve their excellent piece of software.
And the other hand, there are still plenty of bugs in the mkCDrec scripts
too, just look at the SourceForge site for yourself to have an overview
of the bugs which are still open.
From time to time I do get patches from people to solve a problem which
is always nice (see the Changelog of mkCDrec).
Did you have any cooperators on mkcdrec project? What was their
work focused in? Do you still look for some others?
mkCDrec is a tiny project, where I am the main project leader,
code writer, designer, but I do have 2 excellent co-developers on board
which have written and debugged plenty of code. Currently, Mike Flyn (one
of the co-developers) has started a port of mkCDrec to the new Mac architecture
(PPC) using Debian/Linux.
The second co-developer, Franky van Liedekerke, is an excellent debugger
and is still very active on the mailing lists of mkCDrec.
Our plans are to broaden the architectures where mkCDrec can be used
on. I'm thinking of Sparc and Itanium-32/64 platforms. I have an old Sparc20
at home which will be migrated to a Linux/Sparc soon.
The main problem we face is sfdisk which is basically an Intel locked
program. We need to rewrite parts of it to get it working under PPC, Sparc
and IA-32/64. On all those platforms fdisk is available, but that doesn't
give us the flexibility that sfdisk has.
I need to start a sub project to port sfdisk to PPC, Sparc, IA-32/64.
I'm still looking for programmers to help us with this. I do think this
is a perfect subject for a thesis or master-work.
What new features can we expect in the new releases? What are you
planning to implement and in what timeline?
In the upcoming mkCDrec v0.6.8 you can expect a choice to force
DHCP support as requested by users to do mass cloning. More hardware RAID
controllers will be supported as these are quite easy to add due to our
modular design. We are also planning to add support for DVD.
We do hope to have a breakthrough for a PPC mkCDrec version soon.
Furthermore, some small redesign of mkCDrec will be done to build a
real RPM version so the main Linux players can add it to their mainstream
distributions.
My main target is to get a stable mkCDrec version 1.0 out within the
next 6 months. Of course, that does not mean that the current v0.6.* versions
are not stable. I know people who still use mkCDrec v0.5 without problems.
As always, it all depends on the hard- and software one uses.
Surely you have a lot of responses from a happy users. It is quite a
good success, what do you think about it?
In my mkCDrec mailbox I have around 2000 mails. Only 1 mail
reported the loss of a complete system by using mkCDrec without giving
additional comments. So, was it true or not? I will never know! Anyhow,
I always advise to backup your system via other ways before playing around
with mkCDrec, and to use a test system where possible to learn the drill.
Remember I said, mkCDrec is a disaster recovery tool and is not really
a backup/restore tool as such.
Personal I use mkCDrec, on my production systems, once a week completely
aside from my normal backup procedures. When a system crashes I boot from
the latest mkCDrec CD-ROM and restore the system (after repairing broken
parts). Aterwards I restore from backup up to the point of yesterday or
so.
To answer on your question: is it a success? For a tiny project not
bad. I do get plenty of success stories from huge companies using mkCDrec.
That makes me feel good, because those big companies are using Linux in
the first place and mkCDrec just fills up a tiny hole.
Can you describe us your current work, maybe in research and
academic
field. What are you planning to work on in the near future?
I have my own small IT company busy in the field of UNIX (HP-UX,
Solaris, Linux) doing design, system administration, clustering, firewalling,
VPNs, etc... I'm constantly surfing the Web for new ideas for doing things.
And, I like to read and study on computational topics.
In the near future I will start the port of mkCDrec to Sparc. And I
do hope to find the funds for an IA-32/64 machine to port mkCDrec to there
too.
For months I wanted to write a mkCDrec clone based on afio and tapes
to build an one button disaster recovery tape (already done the proof of
concept), but due to lack of time never got around it. The last days the
idea came back and finally started to read how to use afio in-depth.
Maybe, over a few weeks a new project is born much smaller than mkCDrec.
Do you use linux at home? If yes, for what? Please describe a little
bit
I use Linux for years already at home. I use it right now to
answer all my mails, to do research, to do coding, testing, etc. I must
admit I have a Windows 2000 system, but that is because my customers have
it too. The good news is that more and more I can convince my customers
to go for Linux, for e.g. samba, mail servers and firewalls, instead of
sticking with Windows servers. The benefits are a lower TCO with Linux,
because of the free software, and a better remote (secure) maintenance
is possible. Less visits means less costs too for the customer!
My laptops only have Linux, because I mostly work with Unix systems
anyway and then Linux is the perfect glue for X Windows, samba, network
debugging, etc...
What do you think about current position of Linux on desktop and server
market? What are your views about its future?
Linux will become the number one on the server side I'm sure
of that. However, on the desktop side it will be hard to beat Microsoft,
but it is not impossible. Ten years ago Linux was NOT an option in any
company, but now it is very quickly accepted once explaining the pros of
it against an expensive Microsoft server.
What is your opinion about the OpenSource and through-internet
cooperation of developers? How is it possible that unpayed development
works so well?
I must say that sometimes I don't understand
it myself why there are some many developers doing so much for free. Is
it for the fame, or for doing the right thing? I am sure that it pays back
in the long term as companies do take in account you are working in your
spare time for a good cause! I can ask the question to myself.
In the first place because I needed the project.
And, I believe that with making the sources open the development goes much
faster as user feedback and bug reports is almost instantaneously after
a announcement on Freshmeat.net. Almost two years later it turned
out I was right about it. In a closed source environment mkCDrec would
not be able to do all the things it is capable today. I'm proud of it I
made the right decision to GPL the code.
On the other hand one does not need too much
developers as most people have trouble with doing project management in
a correct way if there are too many persons involved.
I may not forget to mention that SourceForge.net
has played an incredible role into the world-wide Internet cooperation
of developers with their servers and sourceforge software. Of course, without
the CVS software it would also be impossible to work with different co-developers.
As you see, Open Source projects do help each other.
It is funny, I never met my co-developers in
real life! I do think we should meet once on a conference maybe?
Some words to our readers for the finish
I can only wish that many, many people
contribute in their own way to the Open Source movement. That doesn't mean
to write only software, but with translating text for projects, answer
user requests, helping building web site (mine is ugly I know), preach
Open Source projects within your organisation where you work, live, etc.
The main advantage of Linux and other Open Source
projects is that the source is available to look at and to change wherever
needed. Of course, sometimes it gets dirty, but what the heck that makes
life exciting!
Interview taken on November 16, 2002 (via e-mail)
Published by PANDOR/linuxzone.cz: http://www.linuxzone.cz/index.phtml?ids=10&idc=474
(December 11, 2002)
$Id: interview_pandor.html,v 1.1 2002/12/12 05:53:34 gdha Exp $
|