For this second interview we are going to South Africa to meet Buchan Milne.
Can you introduce yourself?
I am a 32-year-old system administrator (“Senior IT Specialist”), who studied to be a Mechanical Engineer, but got into IT while at University. I started my working career at a Linux consulting company (working as a consultant to an ISP, and later on a custom Linux distribution based on Mandrake Linux 9.1 for an OEM). The last 4 years I have been at a Telco, working in their ISP.
When did you start contributing to Mandriva Linux? Why did you choose Mandriva Linux as a distribution and for contribution?
While at University, the lab I was working in started doing commercial work, and could no longer use the site licenses afforded to the University (for the likes of Netware etc.), just at the time they most needed some real infrastructure. I had used Red Hat (6.0) while updating their website, but had been frustrated with the fact every time you wanted to install some additional software (that was on the CD, but you had neglected to select during installation) you had to manually track down all the RPMS (the –aid option to rpm didn’t even exist then).
I had tried Mandrake Linux 6.0 or 6.1 to try out KDE, and when 7.0 was released, which made it easier to install any additional software (via urpmi), it solved one of my bigger frustrations with Red Hat.
We used 7.0 as a file server, mail server, web server etc., but we were growing to the point where managing Windows NT4 usernames and passwords for 30 people across 20 workstations was getting to be frustrating for everyone. So, I implemented a Samba (2.0.6) domain controller on Mandrake 7.1, and joined all the workstations to the domain.
Then, Windows 2000 was released, with better stability than NT4 Workstation, but samba-2.0.x was not able to support Windows 2000 domain members. So, in the December holidays, I decided I would learn how to build RPMS, to allow an easily maintainable upgrade on what had become a production server.
When samba-2.2.0 was released, I built packages for a number of releases of Mandrake Linux (by that time, we had a number of servers running Mandrake Linux, releases 7.1, 7.2, and 8.0), making them available on our web server. The Samba team offered to have them hosted on the Samba mirrors.
I noticed that the samba package had not been updated in cooker, and sent the maintainer an email, including my changes, which he merged into cooker.
As samba-2.2 changed quite rapidly, adding new features (such as winbind in 2.2.2), I kept up building pre-releases and sometimes CVS snapshots to test features that would be useful. However, eventually the spec-by-email method of updating packages grew too cumbersome, and Sylvestre requested an account for me on the build cluster. I was then able to make changes directly myself in CVS, as well as add more packages (to contrib) that I found useful.
I was running Mandrake Linux on the desktop (sometimes dual booting), and some of the analysis software we were running happened to run quite a bit faster under Linux than Windows (on simulations that would run for a few hours at a time). So, I looked at moving to LDAP for storing the Unix account details, and naturally it made sense to want the Samba account details to be in LDAP as well. A tutorial by Vince Danen on the www.mandrakesecure.net site helped me to get an initial LDAP setup working, and I later managed to migrate the samba details into LDAP as well. I documented the process, and wrote a follow-up tutorial for mandrakesecure.net.
From there I started taking more of an interest in LDAP-related software. Experience I gained while contributing to the MandrakeLinux openldap packages assisted me in getting my first real job at Obsidian Systems, to work on an OpenLDAP implementation for an ISP.
What components are you taking care about in Mandriva Linux? Why this choice?
In the main distribution, I currently look after samba, openldap, nss_ldap, pam_ldap, and cluster. In contrib I actively maintain xymon, devmon (for which I am also the upstream maintainer), and take interest in a number of other packages which I have used in the past (e.g. grass, qgis) and packages related to software I use casually (e.g. Catalyst modules, dynamips, tac_plus, rancid).
Can you tell us what is the coming challenge for Samba 4? What will be the concrete consequences for users and companies?
Samba tries to bridge a lot of the gap between typical Unix environments and Windows environments. While some of the underlying technologies and protocols are similar, it has taken a lot of effort to get Windows clients to believe a Unix LDAP server and KDC are actually AD. However, a lot of other solutions depend on the protocols involved. So, while Samba4 may solve the “I want AD on Unix” problem, the absence of the pieces required to either make a Linux desktop integrate well into an AD domain (including “managed desktop” features), or to allow Windows members in a Samba4 domain to see other non-filesharing features (e.g. WMI), will be exposed.
Samba4 also means that there will be more embedded devices (e.g. NASs) which will be able to offer lower-cost alternatives to the SME-targetted Windows Server versions.
For Linux distributions, probably the biggest concrete feature at present is MAPI support (e.g. in Evolution) via openexchange using samba4 libraries.
For companies, Samba4 will present both challenges (e.g. in determining how to integrate the “Unix” teams with the “Microsoft” teams) and opportunities (e.g. being able to use existing non-Wintel infrastructure to provide authentication and related services to Windows desktops without missing out on features).
How do you see Linux evolution on server side? What are the strong point and on the contrary weaknesses?
Feature-wise, I think Linux is doing well, especially in terms of catching up with features compared to proprietary Unix, and current distributions allow more flexible and cost-effective solutions for problems where the answer was previously “Big iron” or expensive proprietary software.. However, the challenge (for “generic” Linux distributions) in making products that provide easily configurable but advanced features available remains. Standards such as CIM/WBEM need to receive more attention, as they would allow projects to maintain configuration interfaces for their own software that could be exposed to other tools.
Another natural weakness is the low penetration into the desktop market, which means that even if Linux servers provide a better solution for managing Linux desktops and equivalent features for managing Windows desktops, the primary motivation for implementing Linux servers over Windows servers is financial. While the current financial situation enhances that motivation, unfortunately the steeper learning curve (for techies) and the “unsupported”/”freeware” connotations (for management) provide too many excuses for not experiencing the freedom a typical Linux distribution provides.
I guess in that regard, the third weakness is one of marketing. In some businesses new technologies aren’t implemented until Microsoft has an offering available, while open-source solutions were available which would have cost less and provided more features. This is often due to the people who approve the budgets not willing to take any risks without some analyst report on what the current “CIO” trends are. I think this is one area where the community can help, publishing more whitepaper-type articles and business-oriented success stories (and fewer howtos where 90% of the content could be replaced with ‘urpmi foo’).
Thanks for your time Buchan and all your hard work in Mandriva distribution!
(Samba Project: http://www.samba.org)
