Jim McQuillan and Ron Colcernian - interview

From LXF Wiki

The gentle clients

Jim McQuillan and Ron Colcernian founded the Linux Terminal Server Project eight years ago. Since then it has become an important springboard for many seeking to gain the power of terminal-based computing at low cost. Linux Format meets them.

In the early days of widespread computing there would be a central server, and users would gain access through a dumb terminal that just relayed ASCII data from the keyboard to the server, and back to green text on black screens. The advent of desktop computers changed all that. But in recent years, new terminal technology has revived interest in the concept of central servers and the benefits they can bring in reduced costs and management. This has given new relevance to the Linux Terminal Server Project (LTSP), set up by Jim McQuillan and Ron Colcernian in 1999 to help people use Linux and cheap hardware to realise the potential of networked computing. Nick Veitch caught up with them in the States to see what almost a decade had done for the project.

Linux Format: I think I first met you in 2001 at the LinuxWorld expo in New York. That was just when LTSP was starting.

Jim McQuillan: Well, we started in the summer of 1999, when we went public with the project. We shared a booth at the 2000 LinuxWorld with Etherboot, and 2001 was probably when we had our first booth of our own. So yes, we were still in our infancy at that point.

LXF: And I guess a lot has changed in the interim years?

JM: The technology has evolved a bit; the basic ideas haven't changed all that much. I think our product has gotten better because it's gotten more mature. We've added things like local device support ­ you plug in a USB stick and the icon appears on the desktop. That doesn't sound like that big a deal because everybody's doing it, but you have to remember that in our case the session's running on a server, and the user is sitting at a thin client. If you plug the USB device into a thin client, the session on the server needs to know about it. We put all the plumbing in place to make that happen very nicely; it's a real elegant, robust solution with none of this hackery that we used to do. It's worked out really well.

And the other thing that's happened that's really helped us is that applications under Linux have gotten to be really good. When we first started there was no word processor, the web browser was Netscape 4.something, which was OK, and email clients and all that kind of stuff just didn't exist. And now we have the OpenOffice.org office suite, we have Firefox and Thunderbird, Evolution and all these wonderful, rich applications. So those have grown along with us, making our job easier. If you installed LTSP back in 2001 when we first met, it was quite a challenge. It was a lot of work, there were a lot of manual steps to go through. And LTSP has evolved quite a bit. The installation is very simple now. You download a little LTSP_utils package that contains the installer, you run that and it installs and configures the system for you ­ it's very simple to set up. Up until now, that's kind of what LTSP has been doing.

A year and a half or so ago we got involved with Ubuntu, where LTSP is now integrated into the Ubuntu distribution. We're working now on getting all the major players together: we've got Red Hat, Debian, Ubuntu and a few others to all sit down and really come together in terms of how thin clients will work on Linux for all the distros so that all the distros can include LTSP, integrated right in. My goal really is just to have thin-client functionality built into the distro. You install your distribution, you turn on your thin-client functionality, you plug in a thin client and there you go. There shouldn't have to be this extra step of going to our website and downloading the code. It should just be there, just like Apache is there and Samba is there ­ LTSP will be there.

LXF: I guess that's easier to do on some distributions than it is on others, at the moment?

JM: It is, and some of that is down to philosophy, some of it's technical issues, but everyone I've spoken with so far is really excited about this. Ubuntu's been working on it for a year and a half, and Ubuntu and Debian have been trading back and forth features that have been growing up in Perl with each other, to include LTSP. The Ubuntu developers and the Debian developers work very closely with one another. So that's really the same implementation. Fedora, of course, will be based on RPMs instead of Deb packages; SUSE will be based on RPMs, though certainly not the same RPMs that Red Hat is using. The goal is that when we're done, LTSP will be installed on our server and the thin client will boot, in the case of Ubuntu, Ubuntu bits. In the case of Fedora, the thin client boots up Fedora bits. At LTSP.org our main job will be to provide that little bit of stuff that makes thin clients work that way. For years we've provided all the libraries and the kernels and all the stuff that the distros are so good at doing, so we're getting out of that business to let the distros do it, and we'll provide the part that we need to provide.

LXF: I was going to ask what differences in the development of distros in the lifetime of LTSP have made to the project.

JM: Traditionally, up until now, when you install LTSP it's all LTSP bits, and that's what we're trying to stop doing. I wrote a spec on this subject; Mark Shuttleworth found it somehow and it's kind of what they've based theirs on, the Ubuntu version. In my spec, I said, "Look: 95% of what we ship in LTSP has nothing to do with LTSP ­ it's just stuff we have to have, because LTSP is really like a distro that sits on top of another distro. So let's integrate it into the distro and ship our 5%. Let us focus on the cool things like syncing your PDA, or really robust multimedia support, or security or things like that."

LXF: What short-term challenges do you think there are for you after the distro integration?

JM: The next major thing, which I really wish we were working on now but because of the distro stuff we're not, is good, solid multimedia support. So you can play videos and audio and have the signals synced up properly, and have 30 or 40 clients on a server all playing either the same or different videos. Let them do what people are doing already in Linux, but let them do it as a thin client user ­ that's our next challenge.

LXF: Is that something that people have been asking for, or is it some personal goal?

JM: Yes, they have been asking for it, and they've come to expect it. In the past we've had to get the attention of people to say, "Look, we need help with audio in KDE" or "Look, we need help with something in Gnome." And I always felt that they were these big products and we were this little guy. Now we're at the same table with these guys. We sit down and we talk with the Phonon guy from KDE, that's the audio stuff. We talk with the GStreamer developers, that's what Gnome is using. We talk with other Gnome developers about things that we need. And everyone's like, "Cool, we need that, let's do that." People realise that LTSP is an important way to put their applications on lots of desktops. I'm kind of excited about that.

LXF: What do you think is the reason behind the recent resurgence of interest in thin clients in general?

JM: People are tired of the work that they have to put into maintaining a Windows-type setup. And really, PCs on everybody's desk, with storage and a full PC ­ that's a lot to maintain. As an administrator, that's a lot to keep running all of the time.

Ron Colcernian: And you put your assets at risk, because employees will inevitably store electronic documents that you've paid your employee to create, ten ollars, thousands of dollars, on media that's not backed up. And with LTSP you get this default, centralised backup of everybody's userspace, and by design you eliminate that whole envelope of risk. I don't know how many times in business you've heard somebody say, "I gotta recreate that spreadsheet" As a business owner, that's maddening.

LXF: Yes: lost time, and lost money.

JM: People are starting to realise that thin clients are a really good way to go in many cases. It's not perfect for every situation: if you're doing video editing or big tasks like that, maybe not, but look at how many people just need a word processor, spreadsheet, web browser and email client, and that's all they need. Maybe they have some legacy app that they need to put transactions into a database. Thin clients are perfect for that.

LXF: I guess that business office user, as they like to call it over at Novell, is obviously what they've designed SLED for.

JM: A great way to put that on lots of desktops would be to use a thin client.

LXF: Why do you think people choose Linux for a thin client setup over anything else?

JM: Rather than choosing Windows? A lot of people are just tired of the Windows game: constantly upgrading their hardware and software. Ron does a much better job of explaining this than I do! The hell that they go through. Do you want to say something, Ron?

RC: What motivates people is that they have gotten tired of Windows. I was talking to a guy today. He said, "Admittedly, I use Windows and I make a darned good living from it" And I said, "Yes, perhaps you do. But let's talk about a single day that you spend dealing with Windows problems. A day that you spend loading and unloading software for eight hours." And at the end of the day he got paid, but perhaps unlike yourself, perhaps, myself and Jim, he's added no intellectual value to his brain. He's not expanded his value to society, he's not expanded his value to himself. I think a lot of people are beginning to catch on to that.

LXF: Do you think that most people make the leap from Windows on the desktop directly to thin-client Linux, or do you think that they tend to take it a stage at a time?

JM: Different things happen depending on the company.

RC: I think the answer to that is the scale of the deployment. A guy came up today who had a call centre, I think he had 75 or so seats. And he had no idea about LTSP. He had made a commitment to move away (for a set of reasons not discussed) from Windows to Linux. And they deploy fat clients with full distros on Linux on everybody's desk. Had they had the benefit of researching the project or being familiar with it, they would have jumped directly to a thin-client deployment, because it just makes more sense. Perhaps part of it has to do with getting the word out: if you're doing more than, say, four or five, some magical number, this is a compelling technology to consider.

LXF: Do you ever find it frustrating that lots of people would use solutions like yours if they only knew about it?

JM: I always wish we were better at getting the word out; more people would know about us, because I think there would be more people using it. I do have enough success stories that I can just think about and be proud of. The fact that everybody doesn't know about it ­ that's all right. I don't lose sleep over that. The people who do use LTSP love it.

LXF: What's your favourite success story then?

JM: The one that gives me goosebumps? There are a couple of them. There's a huge deployment down in S o Paulo, Brazil. Six thousand telecentros ­those are cybercafes sponsored by the government ­ 6,000 cyber cafes, 20 workstations and a server in each one. And the workstations work off LTSP. That's 120,000 terminals spread around the city, and S o Paulo is a huge city, the third largest city in the world. There are people down there who have email addresses but don't have street addresses, because they're homeless. But they can still communicate with their family and friends and whatever. They can still browse the web and do government things, you know: sign up for welfare or whatever they have to do. That one really gets me going.

Some of the stuff that the Shuttleworth Foundation has done in South Africa: they're providing computers to 240,000 students in the schools. That's not 240,000 workstations, but they're serving a lot of workstations in each school, serving 240,000 kids who otherwise wouldn't have access to computers. That's significant: those kids are going to come out of those schools with the same skills ­ many times better skills ­ than some of the kids in the US. Those are the sorts of things that put goosebumps on my arms, that get me going.

LXF: Looking further forward, what would you like to have happened ten years from now?

JM: Ten years? That's a long time. The project is only seven years old. Let's go a little sooner: three years from now I'd like LTSP to just be on all distros. Then I'd like to just focus on the hard parts ­ the fun things. So that people just... it's just natural, they just install Linux, plug a bunch of terminals into it, and there they go. That's where I want to be, ten years is hard to imagine.

LXF: Apart from the multimedia streaming that we talked about earlier, what other technical things do you think you need to solve?

JM: Security is a big issue. We're using NFS, we're using X, we're using DHCP and TFTP to boot up these clients. And there's a real security issue there. It's OK for schools or a small office where you only have 15 or so terminals and the threat of prying eyes isn't that big a deal, but we really can't get into the financial institutions because of the protocols that we're using. So we're looking at things like running IPSec between the terminal and the server, or we could run X over SSH to encrypt the X traffic. That places a pretty big burden on the server, but the servers are getting pretty big these days.

That's kind of an area we really need to work on as soon as we've got the multimedia stuff out of the way. As we attract more people and more distros into this, we'll get a lot more help with that kind of thing.

LXF: I was going to ask about that, because some projects never really grow beyond the core team of the developers that started it, but do you get lots of people submitting code?

JM: I get a lot of patches, a lot of contributions. We've got a wiki for people to deposit stuff if they want to put it up there, we've got an IRC channel for people to come in and chat with us ­ we do all of our development on IRC.

We're all out in the open ­ there's no private development going on ­ it's all on IRC or on the mailing list. People come in, some of them help for a little while, some of them stay long term. Now we've been working with Ubuntu for a year and a half, and that's a whole set of developers associated with that distro who are interested in what we're doing. We've got the Fedora team coming on board doing the same thing...

LXF: I was talking to some of the Fedora guys earlier on and they were quite excited about it...

JM: At the LTSP hackfest in September, Warren Togami is coming to work with us. He created Fedora. He started the project; Red Hat hired him as a developer. He's coming to work with us; he wants to get LTSP integrated into Fedora. It's important to him, so it kind of becomes important to other people too. We're getting some real high-calibre people, it's really exciting.

LXF: What do you think is the best way for ordinary users to help you out?

JM: There are a lot of different ways to help us out. They could write code, they could help us solve some of the problems that we're trying to solve. If they can write documentation they could help us with the docs... Translations would certainly help. We have a translation team that we could certainly use more people to help with. If they just want to spread the word, that's huge ­ I mean, get the word out. If they want to sit on the IRC channel for a couple of hours a day and answer questions as people pop in, that's a huge help. If they want to watch the mailing list and occasionally answer the questions that come in on the mailing list, that's a great way to help too. LXF