Barr: Why did you start LTSP?
McQuillan: We started LTSP to solve a problem for a customer. They wanted 35 new terminals to access an AS/400 and a SCO Unix server. We really didn't want to continue
using Windows, so we decided to figure out a way
to do it with Linux. We'd thought about full PCs with hard disks and CD-ROM bootable systems, then we found the netboot howto and decided that this was it. The problem was, the howto did a great job of explaining how to boot a single workstation, but didn't cover the idea of booting 35 separate workstations. That's when we got to work and created something that eventually became known as LTSP. That was back in early 1999.
Barr: What was your first big success with it?
McQuillan: I'd have to refer you to #1 above. That customer now has 140 LTSP terminals, and they have saved hundreds of thousands of dollars with it, and they keep on saving.
Aside from that success, I'd have to say that what
the Brazilian government is doing with LTSP in the
Telecentros really puts a smile on my face.
Barr: Are you happy with its rate of adoption?
McQuillan: Happy ? Absolutely. I'm amazed at how quickly word spread around. We launched the official LTSP Web site in August of 1999, and I think within about 4 hours I was receiving email from people who were downloading it and offering feedback and patches. Now, as with any open source project, it's impossible to count the number of users, but I feel confident that the number of desktops
served by LTSP is in the millions.
Barr: Will you lose control of the project if Novell moves in?
McQuillan: I don't see how we would lose control. We are actively maintaining LTSP, and based on the conversations that I've had with several people from Novell, they just want
to help make it better. I've had conversations with other distros in the past, and they just didn't seem to get it. In this case, Novell really gets it. I mean, if you want to get Linux deployed on the desktop in any size or type of organization, you have to make it easy to manage. Thin clients with LTSP fills that need.
Barr: Are they talking about paying you to work on the project?
McQuillan: I can't go into detail about that, but so far, I'm impressed with what I'm hearing.
Barr: Before you started the project, what kind of expectations did you have for it?
McQuillan: Absolutely no expectations. All we were trying to do was solve a problem for a single customer. We demonstrated the solution at our local Unix user group meeting, and a friend suggested that we post it on the Internet for others to see.
We felt that we hadn't really invented anything new. We just provided some glue to connect a bunch of existing pieces, like Etherboot, NFS, DHCP, X, and Linux, into a manageable solution.
Probably the most important thing we did was the
documentation, which explained in (hopefully) simple terms how it all works. Then, a community sort of emerged, made up of people from all over the world, faced with the same issues of trying to deploy Linux on the desktop.
We've been pretty stubborn about some parts of LTSP. For instance, some thin client solutions work great as long as you have new equipment with fast processors and lots of RAM. LTSP, on the other hand, works just about as well, even with old low-powered computers. This makes it possible for someone to setup a pilot project for no cost. Think about it -- how hard would it be, if you had to actually buy new thin clients to prove that it worked? Instead, grab any old 486 or Pentium from the back room and use it as a thin client. Schools have tons of those that usually end up in the landfill.