Carl Malamud: Internet Talk Radio, flame of the Internet.
Malamud: This is Geek of the Week, and we’re talking to Mike O’Dell. Mike is the Vice President of Research and Development at UUNET Technologies. Welcome to Geek of the Week, Mike.
Mike O’Dell: Greetings.
Malamud: What does a VP of R&D do when you’re just running a network businesses. This is a…no-brainer, right? You don’t need research and development.
O’Dell: Uh, well I’d like to think of it as sort of chief cat herder. Well actually it turns out there’s actually a lot of stuff to do, surprisingly. Because running the network, there’s a day-to-day real-time component of worrying about the routers and the links and the network management software and so forth. But the other jobs have to do with developing new services. I mean, it’s one thing to think of a new service and it’s another thing to sort of know how to cause that service to exist. Because in networks… One of the things I learned at Bellcore basically was that the services that network provide are in fact… You create them by some kind of configuration change in some of the network elements. Like, you a wire to a port and turn on routing. That’s how somebody gets Internet access.
Or you provide some other kinds of service. You install fax software on a machine and bring it up and make it work. And there’s all these things that have to happen before the service is real, right. So the part of the world I look after worries about that, creating new services.
And then to run a network you actually need a lot of information. And you need it stored in places other than inside the routers. We just started looking at writing down the routing policy for UUNET in an interesting way that you can do computations about, and it’s a pretty amazing. It’s very complicated.
Malamud: Well what type of computations might you want to make?
O’Dell: Specifically you’d like to be able to check your routing configuration against one of your peer networks and see that you both agree that you— You know, that we’re sending you the right NAPs and you’re sending us the right NAPs. And that we agree with that set should be. Some of this is the work that Daniel Karrenberg sort of started with RIPE, with the RIPE 81 route registry thing. We’re sort of working with them on that. But the whole problem of even just writing down such things like that is…
Malamud: Is this simply a matter of preventing routing loops, or are you trying to enforce appropriate use policies and do other forms of policy work?
O’Dell: It’s all those. It’s all of those things. For whatever reasons— I mean, you have bilateral transit agreements with some other network, that’s behind a third one and you know, their routes come to you through a tunnel or something like. A lot of odd things happen because of the realities of the well-controlled anarchy. So uh…well-controlled.
Malamud: Well if it’s well-controlled anarchy, right now no one really governs the Internet. We have an IETF and an Internet Society, and we have all these other groups. But ultimately it’s the operators like yourself getting together and trying to decide how to do things. Is that gonna scale? Are we gonna be able to keep this somewhat informal coordinating mechanism?
O’Dell: Well… I guess if you look at business in general, right. I mean, business relationships in general are conducted like this. Companies have relationships with other companies. And one of the things you do as a business is manage those relationships, and make them work, because they’re…you know, no company does everything, and you have to cooperate—particularly in the Internet. I mean, it’s value is that it does talk to everything. And if you mess that up, then everyone loses as a result of it.
There’s a lot of discussion at least on some from the mailing lists lately about an operations task force or something like that. I mean, there’ve been various straw man proposals floated around about getting together for various purposes like that on routing coordination and things like that. And some of those will become more formalized in the sense that you’ll develop a notation that lets you write it down and that it can be recorded and analyzed by programs, so you can tell whether…you know, two people that think they’re talking to each other are in fact—you know, have very different views of what’s going on.
Malamud: But do we need an operations area to coordinate that for us? Something like a CCITT for the telephone world, or something— Do we need the United Nations to run an operations directorate, or the Internet Society, or—
O’Dell: I’m a, uh… I’m not as raving a libertarian as a couple of friends of mine. But I think you make it in everyone’s best interest to do the right thing, and then let that happen. I mean, again a framework, a place for some of that to live, a place for it hang its hat, is probably good. And the notion that it should happen in proximity the IETF is probably a good thing, because the things that are being done these days are being driven much more by operational reality. I mean, if the realities of a routing policy and so forth weren’t what they are, you wouldn’t need BGP 4, right, to do all of these things, right. I mean, if the world was simple you wouldn’t need address classes, there’d be you know, maximally simple cases there’s only one, right. So, that doesn’t work.
Malamud: Well you worked at Bellcore for a while. Do the telephone companies coordinate their activities? Surely they have similar issues of coordination? Do they coordinate their activities in the informal de facto basis, or do they have some central body that does that for them?
O’Dell: Well they have very formal structures for doing this, mainly because they’re large enough to— You know, they in the business are large enough to attract lots of attention from people who think their job is to you know, mind other people’s stores. But it’s important. I mean, a good example is the international numbering plans that let a phone switch here talk to some phone somewhere else. I mean that’s a case of… You need some place to stand to have those discussions. And how complex or formal that gets there’s some outside pressures of various you know, trust concerns and things like that. But it’s really much more of everybody basically understanding that this needs to work.
So, it will acquire more structure over time. But I think it… I mean, the Internet world is still populated by the free spirits that I don’t think there’s… I think any headlong rush into creating some organizational edifice [laughing]will probably not get very far.
Malamud: Eventually, though, will the Internet be like the telephone company? Will it be this this massive, fundamental infrastructure?
O’Dell: Gee, I don’t whether to say yes or no to that.
I hope it’s like the phone company in some ways. In particular that people come to understand that the Internet is a place to do business. It’s a tool that you can use to do business. It’s a marketplace where you can do business. I think it’s the first real example of a transnational marketplace. It’s a meeting ground where people in Tokyo and London and New York can all sit around the same virtual conference table, and you shrink the world down to a point. And previously you had to be in London or in Tokyo or
So in that sense— In the sense that the telephone— You know, imagine what it was like to do business before there was a telephone. Where you had to send mail—you know, paper mail, or you actually had to go see someone to do a deal. Now, you can argue that you still do in a great many ways, but to order office supplies you pick up the phone when you called Office Whatnot, and they send the stuff over, right. I mean, that’s the way it works. And the notion that before you’d have to go shopping, right, is sort of hard to think about these days. So, in that sense, the impact of the telephone… I want it to be like the telephone.
And in some sense that’s the position we’re in now. Imagine what it was like early on trying to sell a telephone to someone, trying to sell telephone service, right. Trying to explain to them what this thing does that’s so interesting, right. I mean, I think that’s sort of the position we’re all in at this point now. We’re here for the reinvention of the telephone. I mean, you don’t get to be there very often when the telephone gets invented. When something which has the potential to change the way people interact and the way that commerce can be done on a global scale, it’s not very often you get to be there.
So in that case, I hope it is very much like the telephone. and the sense that I would hope it isn’t is that the telephone, for very complicated reasons, is thought about in very niche, narrow ways, right. And recently some of the companies are starting to do interesting services using the telephone infrastructure. But it was so channeled into such a narrow definition for a long time, by then there was this huge mass that sort of kept it that way. That you know, “We define what the phone is and does, and anybody else’s notion about what it could be or could do you know…you know, you’re not the phone company. We don’t care, we don’t have to,” right. And that’s the thing that I would assiduously strive to avoid. And the value of this thing comes from all the things we haven’t thought of, right. In the same way the value of the phone— I mean, before you could buy a phone you had to think of all the things you were going to do with it, right, and sort of justify each one. What a wacko thing to do, right. But you know, people still have to make those same cases.
Malamud: So it’s keeping that interface open and keeping the focus on a general-purpose infrastructure?
O’Dell: Exactly. And the job—you know. I mean, it was interesting, and talking with people at Bellcore. And the difference in point of view, in that they… Some of the folks there, some of the old folks out of the telephony tradition, thought of applications in terms of doing specific things, right. And you know, many times in conversations, I would say, “No no, the application is to simply move packets.”
“Well, but what’s in those packets?”
“I don’t know. I won’t tell you,” right. “They won’t tell you,” right. You can’t infer anything, just move the packets, and everything else will take care of itself.
Malamud: Well that’s no way to run a phone company.
O’Dell: [exasperated gasp] Well… [Malamud laughs] Good. You know. That’s fine.
I mean, I think that’s changing. I think that mindset is changing, but that’s a real mental shift. I mean, there are a bunch of folks there that do understand. But the tradition is so long, and so ingrained. So, that’s important.
Malamud: Mike O’Dell, lately we’ve been hearing a lot about convergence. We’ve been hearing about cable companies, and about telephone companies, and about the explosive growth of the Internet. What’s gonna happen with all three of these different mindsets, these different worlds, converge on the same place? How’s that going to affect the Internet, and how’re we going to affect them?
O’Dell: Well, people talk about convergence… I mean the only thing I know about convergence is that when I used to fix color TVs, you’d take a dot-bar gener— You know, you’d switch the picture tube, you’d hook up a dot-bar generator and get all the hashes to line up so that you didn’t have three different pictures on the screen at once—that’s what I know about convergence.
Um… I don’t know what’s going to happen. If I could tell the future that well I wouldn’t have to do this for a living. [chuckles]
Malamud: When you’re running an Internet service provider—
Malamud: —you’re doing R&D for them. And let’s see some large cable company walks up to you and says, “Gee, we want to do this Internet thing,” what does “Internet” mean in the context of your consumer’s cable TV system? What does that mean? Can they send email from their cable box?
O’Dell: Uh…I suspect it won’t be very long before that’s possible. I mean you hear… Just reading between the lines, and this is pure speculation on my part. You know, the zeroth order of machinery is that yes, you can certainly run data over cable systems or at least modern cable systems that’re bidirectional in various speeds. I mean other folks are doing it. That’s doable. And then you read that people are doing set-top boxes that will run you know, Modular Windows and have processors in them and so forth.
So, the issue here is I suspect not fundamentally technological. In the sense that you know, the boxes will get powerful enough. I think there was an article in EE Times about how Bell Atlantic imagines having a cable box with 1,000 MIPs of performance in it. Sort of…maybe that’s if you include the DSPs to decompress the digital video, but—
Malamud: Should I be able to develop my own box and put it on there and get rid of theirs? Should there be some kind of a requirement that anybody’s box can fit on there, or—
O’Dell: Well, it depends on what “their” means. The cable systems probably have some interest in at least some kind of interoperability spec that says that you know, if you put the thing on the network it won’t blow us up. Because cable systems are actually fairly…creaky analog creatures in a lot of ways. And even when they’re digital they’re still analog. So they have an interest in making sure that you know, you don’t put 110 volts back into it. Or what it has to get bulletproofed to the level that the telephone network is in that sense.
Traditionally the cable people haven’t been big fans of you buying your own box and plugging them in, mainly because they got money for renting that. For large values of “Gee, I have no idea…“ For instance like you know, Apple just announced the TV Mac, which is the 14” integrated Mac, 14-inch color, and it’s basically an LC III inside except it has a TV tuner in it. So you can watch TV and…I don’t quite understand what this gadget’s supposed to be other than you know, of other than an interesting straw man to put out there and just see— Because I think the reality at this point is that nobody really knows. I mean all this is so new, and I think nobody really knows what anyone wants because this is a case of in some sense the enabling technology leading the requirements in the sense that if you simply waited around for someone to write a requirements document on what home information access would mean, then you would sit there for a long time. And I think that’s an example. I think people will try a lot of things.
It will always be important that if the “their” you mentioned was The Internet. That yes, it will be very important for people to always to be able to plug their box into the Internet. Because again, the thing that will make it flourish and thrive are all the new services that nobody involved today can imagine. And the ability to plug that in and make that go will remain critically important.
Malamud: Mike O’Dell, it used to be that life was very simple. There was one network, it was the ARPANET. There was this world of the Usenet out there, but that wasn’t on the map as far as the central bureaucracy was concerned. In a way, your company has made life difficult. Because now we don’t have a single backbone, there’s NSFNET as a transit network, but UUNET Technologies has the alternate transit. And NASA runs a transit network. End there’s one called EBONE in Europe. How are we going to make all these networks talk to each other? You’ve been involved in the Global Internet Exchange, for example. Is the solution for global connectivity in the future?
O’Dell: Well there’s a bunch of interesting straw men being floated, particularly around the folks that’re active in the GIX, about how to do some of these things or what’s the next step. There’s a bunch of interesting ideas, and some of them will probably get tried in the next year so.
But again, I think the reality is that there’s nobody— I mean. You mentioned that nobody runs the Internet. The problem is nobody could run the Internet. I mean, there is no organization that is sufficiently powerful, or could be empowered sufficiently, to direct how it works. So in some sense I think the future is like everything else. At some level, people have to agree to get along. Or they don’t. And if they don’t, you know, bad things happen. And if they do, good things may happen. So I think things like the GIX connection points and people basically understanding that the value they bring to the table—well, that the value that they bring to their customers is not just that you can call next door, but you shouldn’t call the other side of the world. And that’s the win.
Malamud: Well your company UUNET appears to be pursuing two strategies for connecting to the world. On the one hand, your network has always had lots and lots and lots of direct connections to lots of the other mid-levels and regionals. You can get right to ’em.
O’Dell: Mm hm.
Malamud: And so, rather than go through the NSFNET backbone you connect straight to SURFnet or other networks.
The other strategy appears to be this idea of an Internet exchange, the Commercial Internet Exchange—the CIX, or the Global Internet Exchange, which is a piece of virtual real estate, a ring where anybody can connect in.
O’Dell: Right. Well that’s really not that different. I mean in some sense, yes. But UUNET has always been big on bilateral connectivity agreements simply because we started out as a fully commercial provider and…you know, you basically had to put your own wires in, right, because you couldn’t run traffic on wires that other people bought. That’s…the rule. And so we’ve been active all along in doing that.
And basically what these connection points are is that rather than— It’s the…you know… It’s n versus n‑square, in the sense that if you create these nexus points, then basically by pulling one wire to that point, you essentially can create bilateral agreements with six of the people, except you know, you don’t need six wires.
Now, there are network engineer reasons why you may want six wires. If for no other reason than single points of failure and things like that. But I think those are as much… A lot of it really is network engineering. I mean, some of it is geopolitical in some real sense. But I think that’s—
Malamud: Well, as a network provider in some ways I’m your worst nightmare. You get somebody like me who all of a sudden starts putting hoards of data into the network, and everyone starts crossing your backbone and you’re unable to plan because I didn’t tell you in advance I’m gonna start a radio station. How do you deal with the idea that the people injecting the data in the network are out of your control? How do you plan a national network in that world?
O’Dell: Well… That is one of the fundamental problems. Part of it is that the pricing models give you some ability to effect…you know, whether someone is liable to do this to you. But at the same time you basically have to make some assumptions about traffic distribution, and then engineer the network based on that. I mean you watch it. And when you discover that one of your assumptions was wrong you move wires and— And that’s actually not that big a deal. I mean, if you go back and look at the…you know. I think back when BBN was issuing the first network engineering reports and they’d say “Well, we need a link from here to here, and from here to there,” I think a lot of people that read those back then said, “Okay, they moved a wire,” and it was sort of duuuh, I wonder why they did that. But nowadays I think a lot more people are understanding—are coming face to face with traffic configuring. And I think that’s—
Again, there are models that are there. Again, that’s one of the things that phone companies do understand a lot about, is doing some of that. Not necessarily with our traffic distribution, but they are at least comfortable with the scale of the problem. And luckily at the moment we’re growing at a rate that sort of the brain can keep in front of the reality.
Malamud: Do services like World Wide Web and Mosaic…Mosaic being the software, the Web being the underlying data. Did those take you by surprise? Did you expect the kind of rapid growth that we’ve seen in that area?
O’Dell: Um… I think if you had anticipated that kind of service, the fact that it was very popular would not have been a great conclusion to reach. The question is…predicting when dynamite services are going to be thought of by someone, well I mean, you know. I think it’s easier to play the stock market.
Malamud: Did the Web influence your network when that took off in popularity? Did that mean you had to do things to your network?
O’Dell: In some sense… Probably but I don’t know. In the sense that the kinda traffic stuff that we look at doesn’t— I mean, we look at gross traffic aggregation I mean, these links carry gigabytes per day. So actually looking…trying to do very much with anything with very very instantaneous traffic traces… I mean you can look at those and try to say “Ooh, this is Web traffic” or something. The problem is that you can’t look at that I think on a scale that will tell you very much. I think primarily you have to look at the gross traffic flows. Now—
Malamud: What about multicasting? Is that an example of a single service that does [crosstalk] impact you?
O’Dell: Uh, that one yes. That you actually worry about. Because that’s an example of where one stream comes in and several streams go out. So there’s a multiplier effect on that. And in some sense if you don’t do it right, then your customers will get their own tunnels and backhaul the data the wrong way across your network. So multicasting is absolutely something that we think about a lot.
Malamud: Is that because we haven’t fully deployed multicasting in our routers yet, or is it because fundamentally multicasting is just gonna really impact—
O’Dell: No, fundamental multicasting does something the other traffic doesn’t do. I mean, a point-to-point circuit, right, there’s only that traffic. When a data stream goes into a multipoint distribution, it fans out eight ways, right, eight times as much goes out as went in. So that’s a very critical thing. So the ability multiplies the load, okay. And there you actually have to look carefully at where you put the multicast routers, and how much you’re loading links, and things like that. And I think that’s… Again, not many things… There really isn’t very much experience in doing this kind of stuff. I mean, conference bridges and distribution nets that have been done before with the phone folks really fairly different critters. I mean, they’re analog, they’re engineered very carefully one at a time, and they’re not the kind of…you know, what I call splattercasting. You know, you through this thing and it explodes and the data goes all directions. That’s a fairly new gadget. And understanding—you know, trying to think about how you want to do things, and how you think about that is a planning process. I think everybody is grappling with how to get a handle on that.
Malamud: When we do multicasting today we tend to do it in a global sense, which the Internet has always been. I can talk to any other person on the Internet. Is it possible that multicasting will force us to begin thinking in terms of smaller spaces. I mean, might UUNET run multicasting and just do events within your network instead of just everything go out to the world at the same time?
Malamud: Are you gonna be a TV station, in other words. Is UUNET gonna have to get in the business of scheduling and disseminating content.
O’Dell: Oh, gee. I would— Just sitting here thinking about it for ten seconds I would hope not, simply because I’ve got enough to do with what we’re doing.
No, I think part of what we’re seeing is part of the state of evolution of the current multicast stuff. I mean, Steve Deering’s new stuff that does the pruning so that in some sense— You know, rather than send all the data everywhere all the time, you can basically send a connection back up— The spanning tree is created dynamically, and you can send a request back up and the data tees off sort of at the best place. And I think that this is an area that’s really ripe for research. And I think the real improvements will come in fundamental understandings and fundamental algorithm improvements on how to do it. I mean, this is one of the problems that I think is you know, fundamentally hard with a capital H. And that it just requires serious brain power to get it better. So…
Malamud: Mike O’Dell, you’ve bridged two worlds for a long time. You’ve actually bridged several worlds. You worked at Bellcore, yet you’ve done a lot of Internet work. You’re active in the IETF, but you’re also on the USENIX board of directors and you’re the editor for Computing Systems for USENIX. Are these different worlds, or are they coming together? It used to be oh I do Unix, oh I do networking, oh I do databases. Is this now one big technical community?
O’Dell: Yes and no. There certainly are…there are large intersections. And some of them unfortunate, like somehow this rumor got started to you know “Internet is something Unix machines do.” With it with the reverse implication that only Unix machines do Internet, right.
Malamud: And that’s what I was leading up to.
O’Dell: Right. That’s—
Malamud: I mean, is the PC world part of the Internet?
O’Dell: Oh absolutely. Absolutely. I have a Macintosh on my desk, that speaks TCP. I mean, and I can telnet around just like it was some other kind of computer. So—
c Are you running Unix on it?
O’Dell: No. It run runs MacOS.
Malamud: But how do you get a command prompt?
O’Dell: [laughing] Uh… I have an X Window Mac [indistinct] Unix machine.
Malamud: Well thank you very much.
O’Dell: That’s for when—you know, when you itch, right, and you just have to type something.
Malamud: We’ve been talking to Mike O’Dell, and this has been Geek of the Week.
Malamud: This is Internet Talk Radio, flame of the Internet. You’ve been listening to Geek of the Week. You may copy this program to any medium and change the encoding, but may not alter the data or sell the contents. To purchase an audio cassette of this program, send mail to email@example.com.
Support for Geek of the Week comes from Sun Microsystems. Sun, The Network is the Computer. Support for Geek of the Week also comes from O’Reilly & Associates, publishers of the Global Network Navigator, your online hypertext magazine. For more information, send email to firstname.lastname@example.org. Network connectivity for the Internet Multicasting Service is provided by MFS DataNet and by UUNET Technologies.
Executive producer for Geek of the Week is Martin Lucas. Production Manager is James Roland. Rick Dunbar and Curtis Generous are the sysadmins. This is Carl Malamud for the Internet Multicasting Service, town crier to the global village.