December 1998
Navigate    Home    Topical index    Archive    Discussion Forums  Subscribe    Free e-mail alerts  Search          Contact Us    Masthead    Advertising Info    Writer Guidelines    Link to LinuxWorld  What do you think?    E-mail the editors

Advertisement

Interview with Eric Allman

Commercial development good for Sendmail Eric Allman reflects on the release of his Sendmail Inc.'s first commercial software Summary Earlier this month Sendmail Inc. released its very first commercial product -- Sendmail Pro. LinuxWorld senior editor Robert McMillan caught up with sendmail creator (and CTO of open source software company Sendmail, Inc.) Eric Allman recently to find out how life as a commercial ISV was treating him. Though he was frantically preparing for this month's product release when we spoke with him, Eric said that things were good and getting better -- both for him and for sendmail's users. (1,100 words) By Robert McMillan

LinuxWorld: So what's the plan with sendmail? Do you continue to develop a core free product and sell enhancements?

Allman: Well enhancements are going to move into the core product as well. the Sendmail MTA (Message Transfer Agent) open source is not a dead-end product. If anything it's got more life now than it did before because we can put a lot more resources into it.

Eric Allman

I was in the position where I needed to need a living. I went out saying "OK I'm going to be a consultant." I figured "I'll be able to work on Sendmail full time. I'll take some consulting jobs. Surely people will want to pay me to do this..."

No. It doesn't happen that way. What happened was Sendmail development ground pretty much to a halt. I was earning a living and Sendmail got to be so popular that I was spending all my time maintaining the old version, not working on new versions.

LinuxWorld: After you decided to form Sendmail, Inc., How easy was it to sell your business plan to investors?

Allman: We had to explain this very foreign concept. You know, "Gee, we're going to give away the core product and we're going to make money at it." This was one of the big things we had to explain clearly in the business plan, and it wasn't easy to get across. People are used to selling product, not giving it away.

LinuxWorld: Do you think this has changed at all?

Allman: Clearly it has for us. Investment psychology is really kind of interesting. Once you get a certain number of people signed up, the perception is that risk has gone down and other people want to come in.

For us [the key] was getting Bill [Joy] and Andy [Bechtolsheim]. In all fairness, Bill did enough with Berkeley Unix, which is an open source model. He didn't need a lot of convincing. Andy didn't need much convincing either. And once they were on board, you know we suddenly had these high profile investors, so then suddenly it became easier to get other investors.

LinuxWorld: How has the reaction from your user base been?

Allman: So far it's actually been fine. We'll see what happens when the commercial product comes out [Sendmail Pro was released this week]. There's still a possibility of something there. But it's actually been almost completely positive.

The first couple of weeks, a few people grumbled about it and then they just kind of went away.

LinuxWorld: I imagine the first thing most people thought when they heard of Sendmail, Inc. was that they were going to have to pay for Sendmail. How did you reassure your users of your good intentions?

Allman: It didn't hurt that when we announced the company, we announced the company and "oh by the way, here's Sendmail 8.9. It's open source. It's free. Take it." That's a pretty substantial statement.

If we end up releasing a commercial product and the traditional open source types say "Hey that's a pretty cool GUI after all. Maybe I could use it, but why hasn't he open-sourced it? Gee, Eric is just a money-sucking scumbag; he should give it to us." Then that's going to be a serious problem.

There has been absolutely nothing whatsoever that has prevented the open source community from coming up with the GUI for Sendmail. No one has.

I don't think it's impossible that we will release the GUI at some point, but it's not going to be immediately. We have to recoup our costs and make some money for our investors. What I know, though, is there has been absolutely nothing whatsoever that has prevented the open source community from coming up with the GUI for Sendmail. No one has.

Why is that? It wasn't sexy enough, or what? I don't know why they haven't done it. A couple of people have done some stuff that was mediocre. They didn't really address the real problem; they sort of took the text configuration and put a simplistic GUI around it without thinking about it the hard way.

I think that this actually points out one of the failures of the open source model. People want to work on the neat stuff. And the problem is there is a certain amount of scut work you have to do to make a complete product. People don't like doing scut work.

Sometimes it's good to just pay people. "Yes I know it's boring work, but here's your paycheck. Do it anyway." You can't do that in an open source model. So what happens is a lot of stuff that really ought to be there just doesn't get done.

LinuxWorld: So what other features does it make sense for Sendmail to develop for the commercial release?

Allman: The sorts of things we'll probably look at in the future, are the types of things that are probably not of interest to the broad open source community.

I don't want to talk about the specifics, but large corporations, large ISPs (Internet Service Providers) have certain kinds of problems that other people don't have.

The Unix mailbox format, where one file contains all the mail for a given user, actually works OK for a small number of users, but as you start to get into a large ISP size, where you've got a million mailboxes, it doesn't work anymore, because they're all in one directory and you can't put a million files into one directory and be able to search that directory fast enough. So you have to start looking at other ways of storing the mailboxes.

This is not of interest, except to the top 0.5 percent of sites. It's not clear that it makes sense for that to go into the open source product. But for the people that need it, they really, really need it and it costs them a lot of money not to have it.

LinuxWorld: What kind of enhancements are you looking at for the open source release?

Allman: An example of stuff that probably will go into open source: There's an SMTP extension called enhanced status codes. The SMTP server says "In addition to the standard SMTP reply code, I'm going to hand you enhanced reply codes too." So this could be useful, and you want as many people as possible on the Net to be running it.

We'll put that into the open source. It's not one of those incredibly sexy things, but it's the right thing for the world and we'll get it in. That's also the sort of thing that I probably wouldn't have had time for on my own.

Yes indeed, there is an infinite list of those kind of things.

LinuxWorld: How many engineers are working on the open source enhancements right now?

Allman: We don't have engineers dedicated to the open source. As we do things, we make a decision whether that feature is going to go in. Everybody's working on the open source, but not 100% time.

LinuxWorld: I'm trying to get a quantification of the open source work. It seems significant, but...

Allman: I don't have it for you.

I'll be honest with you. The majority of what we've been working on to date has been getting the GUI together. We've got to get our first product out. We're a small company, and we're still hiring engineers about as fast as I can.

Discuss this article in the LinuxWorld forums (2  postings)
(Read our forums FAQ to learn more.)