Editor's note: At NetWorld+Interop in Las Vegas, Network World sponsored a showdown among vendors of network-based operating systems. What follows is a transcript of the showdown. Unfortunately, audio problems mean that some of the statements were not transcribed.
Good morning. I'm John Gallant, editorial director of Network World and I welcome you to this Operating System Showdown. I think you will find this a very lively debate and - I hope- informative session today. I hope this carries more light than heat. But if it generates a lot of heat, well, that's okay too. It will be entertaining at the very least. But I hope we get some good answers to your questions here.
This is, if you haven't attended one of our showdowns before, we've put on about a half dozen of them. These are real debates focusing on key issues like network and systems management, thin clients, Layer 3 switching. We've done all sorts of things.
Today I think is one of the more important showdowns that we've done with our debate on operating systems. And I think you'll be pleased to see that we have the leading operating system vendors up here today; from Microsoft and Novell to Red Hat, the leading commercial Linux vendor, to SCO, the leading Intel-based Unix vendor, and Solaris Sun up there representing the RISC Unix community. So I think we've given you a good range of vendors here.
And I'll introduce our panelists for you in a second along with our panel of experts, which is evolving as we speak right now. One of our panelists is not here; Nick Petreley. So Rick Villars from IDC is filling in for him. And I think that's wonderful, Rick. Thank you.
So again, no Powerpoint presentations today, no formal speeches beyond an opening statement that they'll get to shortly. It's a real debate. And I'll go through the format in a second. They'll be taking from this panel of experts here. Then they'll be asking each other questions. And finally near the end of this - we have 90 minutes slotted but I don't think we'll go that long - you'll be able to ask questions. We've got some microphones here, or you can stand up in the audience and ask the vendors questions.
We don't allow any group questions. Pointed questions to specific vendors. So I think you'll find that interesting.
I think the timing for this is great. There's just so much going on in the operating environment industry right now. There's so many changes going on. And customers are at a point where they're making critical decisions about where to host their e-commerce applications as well as their critical line of business applications. It seems like every time we turn around some significant change is going on in the industry Microsoft is working on the release of Windows 2000, one of the most heralded software projects certainly of this decade. Novell has just released a new version of NetWare. And certain under CEO Eric Schmidt the company is revitalized.
I would wager that like me a couple of years back not many of you were thinking of Linux. But we have seen companies like Red Hat really come out of the gate strong in the commercial Linux market. SCO and other vendors are working on new initiatives along with Sun and other companies in the Unix community. So there's just a tremendous amount going on, and now is a perfect time to learn about these things as we're awaiting some of these new initiatives and as some of these products are taking out in the market.
Again, our goal here is to give you information to help you make more informed purchasing decisions. So I really hope you'll take advantage of the opportunity to ask these vendors some questions. It's very rare that you get the chance to see five technically astute people like we have up here this morning and to be able to ask them questions while they're all in one place. So don't pass that up. And I also have poker chips from the Hilton to give out for really good questions, but you don't need that incentive, right?
Let me introduce our panel of experts. Closest to me is Rick Millers with IDC. Rick, your title is?
RICK VILLARS: Vice President.
GALLANT: Vice President of IDC. And Rick, again, is filling in for Nick Petreley, who is the editor of Linux World and helped us prepare for this. He must still be in a plane circling over Las Vegas in a sandstorm or something.
And next to Rick, we have Christine Burns, who is director of Network World's Enterprise Test Alliance. And she is one of the top reporters on operating system issues.
Now, our vendor executives-if you had seen some of the advance stuff, we have a change in the speaker from Microsoft. Our speaker today is Jim Ewel, who is Director of Marketing for Windows 2000 at Microsoft. Jim, Welcome.
Drew Major, chief scientist for Novell. Drew, welcome.
Erik Troan, Director of Engineering for Red Hat.
Tamar Newberger, who is Director of Server Product Marketing for the Santa Cruz Operation.
And finally Brian Croll, who is the Senior Director of Solaris Product Marketing for Sun.
As I mentioned, I would walk through the format with you. We're going to have 90-second opening statements from our vendors. And we have a very strict time limit. There's a person down front that you probably can't see who actually has a timer and a little light that administers an electric shock to the speaker [laughter] when they get to the end of their time limit. And that's probably one of the most fun things during the debate.
After the opening statements from the vendors - the 90-second opening statements - our panel of experts are going to go through and ask questions. The vendors do not know what questions are going to be asked; nor do they know what questions they'll ask each other when they get to that portion of the debate. The vendors will have 90 seconds to answer the question. If the questioner doesn't feel that the question was answered adequately or warrants a follow-up, they can ask a follow-up, and then there's a 45-second answer to that follow-up. I'll try to make sure that all the vendors are asked the same number of questions in each round.
Then we turn to the vendors, and the vendors will ask each other questions. And as we've seen in earlier debates, this part is the most fun. The vendors really prepare for this, and it's nice to see.
Finally we'll have a lightning round where the vendors can ask anyone they want a question. During that first part I'll make sure that all the vendors ask and answer the same number of questions. But after that if they all want to pick on Tamar, they can, or any of you. Using you as a representative.
And then we'll take questions from the audience. And, again, like the questions that the experts asked and the vendors asked. We want specific questions as opposed to group questions. So if you could write those down and prepare, that would be great.
And finally we'll get a brief summary of the event from our panel of experts here; what they thought about it and what lessons they took away from it.
So without any further ado, Cheryl, are you ready on the timer? All right. Let's get our opening statements. Let's start with you, Jim, your opening from Microsoft.
JIM EWEL: Well, thanks, John. When I got signed up last week to appear in this I was a little apprehensive at first. You know, people said this is going to be a dogfight, and here I am with my major competitors here. But I'm actually very excited to be here and to have this opportunity to really get some of the facts out, because I think you as a customer want to see those facts in making your decision.
We're about to launch one of the biggest products in our history Windows 2000 server, and it is a very important product both for us and for the industry. People ask me a lot what is the product about. And really I think it's about three things that I think are very important to you as a customer. It's improved reliability. It's about lower cost. And it's about applications.
We've done a lot of things in Windows 2000 server to improve the experience that you have. For example, there are 50 scenarios where we no longer have to reboot the server to do things like add network cards, and some other things that the customers wanted to do.
In terms of lower cost, simply we've added Active Directory, we've added Intellimirror. We've looked at the complete end-to-end picture of your organization and tried to concentrate on not just lowering the server costs, but also lowering the entire cost of the network, including the clients.
And the last piece is applications. Customers don't buy operating systems. If I gave you an operating system under your tree for Christmas, I mean, what would you do with it, right? Customers buy applications. And NT is about enabling great applications.
So those three things; greater reliability, lower cost and better applications.
GALLANT: All right. Thank you, Jim. I think you were wise to be apprehensive.
JE: True.
[Laughter.]
GALLANT: Let's take the opening statement from Novell.
DREW MAJOR: Okay. It's a pleasure to be here to talk about something I've been coding for now 18 years. NetWare is different than any of the other OS's here. We're a special purpose platform that's focused on networking and the requirements of networking. We have a design center that's different than a general purpose OS. And so in a sense we're a kind of in-between IOS, which is Cisco's OS, and a general purpose OS.
We are a very fast protocol engine. We're very efficient in managing hard disks and sell solutions that are built around those requirements; like file service, Web caching, et cetera are best run on our platform. We also have the most widely deployed and most robust directory in the industry. And we are emerging as the best platform for Java applications.
NetWare is extremely efficient in how it uses hardware. Customers regularly tell us it requires two or three times the amount of hardware to deliver the same solution on top of a general purpose platform like NT or Unix. And the benchmarks back us up. Netbench we typically-which is a file server benchmark-we're typically 50 to 100% faster. In Web caching there was recently an event - the first industry standard caching, Web proxy caching event, sponsored by [the National Laboratory for Applied Network Research]. We literally blow everyone else out of the water.
Many of the vendors, including ones here, declined to show up. And then some that showed up chose not to publish. We were literally hitting 1,500 - we showed 1,500 hits per second; we have since maxed out to 1,800 hits per second. Some of the other products were around 50 to 100 to 200. And that's a combination of the unique architecture of NetWare as well as the other stuff that's underneath it.
So anyway the combination is we're different, but there's a lot of value in what we do.
GALLANT: There's typically a breaking-in period where I get people used to those times. Eric, take it away for Red Hat.
ERIC Troan: For me I guess it's a thrill even to be up on this stage to see Red Hat acknowledged and accepted as a leading provider in the Linux operating system space. It's really exciting for me. I've been with the company a long time. So this for me is just really pretty neat.
Now, the message that Red Hat is really trying to bring to our customers is that openness is important. Open APIs are important so that your applications, and your in-house applications especially, can move between platforms and you can avoid vendor [lock-in]. Open source code is important so you don't have to rely on the vendors to fix the bugs in your mission critical applications.
Just now network-wide ERP solutions, running Web servers that have to be up 24-7. If there's a bug and you can't get it fixed, who are you going to call? You call a vendor and pay him a lot of money and hope that they happen to care about you because you're a big enough customer. With Linux we give you the source code. You can fix it yourself. You can choose to buy a support contract from a vendor like Red Hat or many of our competitors. You can go hire a consultant. And many big companies have the IS talent to fix their own problems.
And we're sure by giving you the ability to move your applications between platforms an deliver reliability by giving you the ability to fix your own bugs and avoid vendor lock-in, that we're really delivering a value that nobody else can [inaudible]. We [inaudible] about our performance if you can really take advantage of hardware. We run a wide variety of hardware and we're excited about all of that. But avoiding the lock-in and giving everybody the choices that they need to run their business [inaudible]
GALLANT: Very good. Thank you. Tamar, let's hear what you have to say from SCO.
Tamar Newberger: You know, when I told my grandfather I was going to be speaking about Unix in Las Vegas, he said, "Really? I didn't know Liberace is still alive." Anyway, no one- let's make it simple. No one ships more Unix servers than SCO. Our platform is Intel and we have over 80% of the Unix and Intel market. You might say that we take the risk out of your business. Our flagship operation [inaudible] It's the fastest-growing commercial server operating system on the planet for the last two years running.
Let me tell you a little bit more about Unix 4.7. It's the most advanced Unix operating system on Intel. No one beats [inaudible] square when it comes to reliability or scalability. We're the only ones who can support platforms from a small [inaudible] devices all the way up to complex data center operations on a single platform.
What we believe in is network computing, which is a server-centric clients and hardware independent model. With Unix 4.7 Webtop, our latest innovation, any legacy current or future application can be run from any client anywhere anytime.
Because we own the Unix and Intel space from a 32-bit market, our partners are not joining with us in Project Monterey. With IBM, Intel, Compaq and others, we will be delivering the strongest partnerships for 64-bit solutions.
At the end of today you will find you have a choice between Unix ware NetWare, shelfware beware and nowhere.
[Laughter.]
Thank you.
GALLANT: Okay. I think you're going to have to come out of your shell if you're going to do well [laughter]. Well, Brian, how would you like to follow up now?
[Laughter.]
BRIAN Croll: I'm going to start out with a quote that in fact you might find interesting. It's in today's paper. "In the Internet world the PC will be central. But other devices will become more important." Isn't that an interesting quote? We might think that Scott [inaudible] said that. That came from Brad Chase, who is Vice President of Microsoft's Consumer and Commerce Group.
The point to this that's very interesting is there is a lot going on right now in the software industry. And the reason why is because we're going through a fundamental change on how computing power is going to get delivered to consumers.
This new model looks a lot like a utility. As a result when you look at Solaris, what we're doing is we're building the foundations and the building blocks for that new utility style computing; obviously built on the Internet.
Now, to do this, what you've got to think about, if you're going to have [inaudible] devices hooked up to this infrastructure. This is a challenge that nobody has even attempted to tackle before. As a result, when you look at the servers, the software underneath that infrastructure, you have to hit a level of scale that's never been seen before; a level of reliability that we've never seen before in computing. And you have to have availability of those computing resources all the time.
With Solaris we think we have the best without a doubt foundation on which we can build this entirely new model of computing. This is what [inaudible] Solaris and this is what we're going to be doing.
GALLANT: Very good. All right. As you can see, we requested that they send their shy self-effacing types to the debate today. That's the last controlled piece of the program today. That's the last piece anyone has control over. So we're going to turn now to our panel of experts. We'll start off with Christine. Just let me remind the vendors, on the first question a 90-second time limit on the answer. Christine, Rick, if you feel that you need a follow-up, feel free to toss that out. And it's 45 seconds on the answer to the follow-up question.
Christine, the field is wide open to you.
Christine Burns: I'm going to hit on Microsoft first. I want to know if you've done any testing on the effect of Windows 2000 on the network infrastructure. You've got at the directory managing everything from your users with Intellimirror to your router, so you're going to have to-what kind of servers are we going to need to support this, what kind of network gear, and what are the bandwidth restrictions?
JE: Windows 2000 server is probably the most heavily-tested product that we have ever shipped. And in doing the testing the two important ways in which we have done things differently - one is in terms of the internal tests. We've done a lot more code review and we've done a lot more long-haul stress tests. This actually runs where we run for four to six weeks and we run tests that really really stress the operating system; every build that we do it.
The second thing that we are doing to make sure we really understand how it affects the network infrastructure is that we have a program called the "Rapid Deployment Program," where we have, oh, I think there's about 50 or 55 customers in that right now that will roll out that product in production before we ship it. In fact, we're committed that we will not ship the product until we have customers in production and we understand exactly how it works in their real world environment.
CN: So what about your server requirement?
JE: Are you asking what the specific server requirements will be? Well, on a single processor server I think the box will probably say 128 meg. But in a realistic environment for most customers it's probably 256. On a four-processor server people will want ore memory than that.
It depends a lot on what you're using it for. If you're using it for a file and print server you'll need less memory. If you're using it for a line of business application like SAP or so forth, you'll need more memory than that.
GALLANT: Okay. Thank you, Christine. All right. Anybody but Jim from Microsoft can take a question from you now.
GALLANT: Okay. I'll be working here a little bit without a net, but I'll start with Eric. I'm going to take us on a little bit of a trip down memory lane. Creative Labs just released a Linux driver for Sound Blaster, the sound card. And the problem is that it's a closed source binary [inaudible] product, and as such it only runs on a couple of versions of the Linux kernel. This method of doing it was bound to happen; it's happened in the past in other operating systems. And we have to know about it's not the way that Linux is supposed to be. If [inaudible] was here right now, what would you do to them to convince them that they need to move to [inaudible] to see a more open path?
GALLANT: You know, [inaudible] business really isn't up to me. I wouldn't really try to talk him out of doing. I would point out to him that there is going to be open source drivers for sound cards. Those drivers are going to work better. They're going to be better integrated with the [inaudible]. They're going to deliver the customer more choice in what version of the kernel they run and what sound card they run.
And I would ask them do they really have so much intellectual property and DSP [inaudible] that they feel like they can't share it with the world? [inaudible] be up to them. I think if [inaudible] choose to do closed-source drivers for [inaudible] they're going to end up with sub-par drivers that don't work as well. When you look at the little SCSI people like Milex and Adaptec, they're actively supporting Linux driver efforts to the point that Red Hat has Adaptec drivers for cards that aren't yet released, running. And Adaptec does that because they know that the only way they're going to get a high performance reliable driver for Linux is to deal within source. And I think that applies equally to Sound Blaster and the other sound card out there.
GALLANT: Rick, follow-up?
RICK NONES: Nope. That's all.
GALLANT: Okay. Christine, Drew at Novell, Tamar at SCO, Brian at Sun eagerly await your query.
CB: Okay. Drew, you said yourself that NetWare is a specialized networking operating system. That could relate to your problems of getting accepted as a [inaudible] server. You put all your bets on server-side Java. But is there a Plan B, since service site Java seems to be stuck in the gate, so to speak?
DM: There's really a Plan B. There's a continued focus, though, on solutions around the Internet. In caching - you look at the audio/video stuff coming up, directory. There's tremendous value for just having more special purpose solution. I think that's shown up most currently by what we're doing with appliances. We're releasing a caching appliance with Dell and Compaq that is a special purpose solution. But it does cache extremely well and it is cost competitive. And I would see, if anything, our evolution is more towards devoting more and more of our services as appliances because that's how more and more customers want to deploy them. And so in a sense we evolved from [inaudible] into an OS into a platform for appliance services. And we see a lot of growth there. And let the rest of these people fight the app server battle. We'll fight it with Java and we'll deliver network services and increasingly as appliances.
CB: So if server-side Java doesn't come to fruition, you're not going to push in the app server market?
DM: I would say it depends on the class of app server. Oracle does very well on our platform. Certain other things. But, again, we will be different. I'm different. Do people run apps on Cisco? You know, it's just a different focus. And I believe given the class of needs and how the Internet is evolving, there is a need for a platform that's positioned in a sense in between the routers and the app servers, and that's us.
GALLANT: Okay, thank you. Rick, I see Tamar and Brian down there kind of looking off in the distance. And I think it's time to bring them into this event. Would you get one of them a question, please?
RN: Yeah. I'll actually go with Brian. Sort of an old legend in the Unix industry, and I think it's used against NT is the phrase, or an old saying that the reason people buy so much NT is that they have to deploy it to support the same amount as one or two Unix servers. Do you really have benchmarks to go to people and prove to them that that's the case?
BC: Absolutely. If you look at almost any benchmark up and down from Web servers to application serving benchmarks, I/O benchmarks, number of CPUs supported, up and down almost every single possible measure of skill [inaudible] I think we do very well on. I think this is largely due to the core of what Solaris is all about. It's wholly multi-threaded. It takes advantage, as I said before, up to 64 processors. I mean [inaudible] which is almost like magic. It's a pretty amazing thing. If you [inaudible] almost across the board you want to look at linearity as you add CPU's to the system, Solaris is going to be great.
GALLANT: I guess the follow-up from my perspective is, again, moving beyond a word processor or a couple of processors to a cluster environment, have you done the benchmarks on that side to set up how you're going to work?
DM: Well, actually, yes. There is a clustering product called - a parallel database product available on Sun where we get excellent clustering skill ability through multiple systems being applied to an Oracle database, for instance, as one example.
GALLANT: What's going to actually be interesting is as most people today buy clustering for reliability and availability, typically you can get scaling out of just performance if you do it right out of larger scales; more CPUs, eight-way, so forth. So what you find is the market really is looking for availability out of clustering. It's only when you can't deliver scalability through the software and the hardware that you would move to clustering really as an alternative for scaling.
GALLANT: Okay, very good. Tamar, we haven't forgotten you. Christine. Christine, SCO.
CB: The Linux/NT debate gets a lot of press. But if you do an apples/apples comparison, Linux probably competes better with SCO Unix. Why would I pay your prices to get something that I could get for free from a commercial product?
TN: Well, the problem with your question is that, of course, you can't get the same thing for free. If you could, that would make a sensible point. [inaudible] first of all that we - as you say, we embrace the Linux [inaudible]. We like the fact that it validates a Unix operating system on Intel. We like it so much in fact that we have Linux compatibility - Red Hat Linux compatibility, in our Unix 4.7 operating system. So like the model people are developing on Linux and deploying reliable commercially available backed servers like Unix Ware.
In particular more interesting is - certainly I wouldn't have a technical head-to-head comparison here. But in areas of scalability we [inaudible] off of Linux [inaudible] reliability [inaudible] sort of things like clustering. And then, of course, there's an issue of the road map. The road map, I think it was for our customers one of the most important things. [inaudible] actually say, "Where are we going in the next 18 months and what to deliver on that road map." And we also had this infrastructure as support to back that up. And I think that's really [inaudible].
And I want to remind people, since my [inaudible] gone off, that we [inaudible] free Unix Ware product where people can get for educational and non-commercial use to use, and the same for a source code license.
GALLANT: I suspect that Eric would like to ask you to follow up. And I'm going to ask Christine [inaudible] follow-up.
CB: What are these performance numbers that you're [inaudible]
TN: [inaudible] scalability obviously Unix [inaudible] the Data Center edition just released runs and is [inaudible] not only runs but is [inaudible] for 16-way SNP machines, which is not [inaudible] that Unix has. We have a clustering solution that goes up to 32-way and beyond. [inaudible] shipped by Sequent runs on 32-way SNP machines. These are not numbers that, of course, Linux has at the moment. We have a clustering product [inaudible] on Linux. Our large memory support, including our specially [inaudible] databases which gives us huge [inaudible] performance databases [inaudible] for performance numbers [inaudible]
GALLANT: Okay, very good. Rick, the field is wide open to you again.
RN: Okay. I think I'll go with Jim. Easy question here. Is it true that Microsoft is still using AS-400 just to run some of your machine-critical apps?
JE: No. That was an easy question. No. That data is about three or four years old. And we actually run our business on NT. For example, we run SAP. That's run on an NT server with SQL Server 7. All our financials are done there. We run, you know - a big data warehouse is on NT and SQL. The only things we have that aren't NT and SQL, quite frankly, are some acquisitions that we've done, where we'll gradually move them over. But we don't have anything in the [inaudible] for a hundred other than some testing for SNA Server.
GALLANT: Okay. And now actually that brings me to the follow-up then. Let's talk a little about one of those acquisitions; Hotmail. Can you walk us through a little bit of the things that have been going there as you have tried to look at NT in the Hotmail environment?
JE: Yeah. When we acquired Hotmail, they were in fact running on Unix. And we have gradually been converting them over to NT.
[Laughter.]
So, for example, we converted the databases from Oracle on Unix over to NT and SQL. And you know what? We're getting a lot better performance on lower cost hardware. Now, we haven't converted all of the mail servers over. A lot of that has to do with - they have a lot of very custom code that was written for that system. And we are going to gradually port that over to NT. But you're perfectly correct in pointing out that we have not yet.
GALLANT: Okay, Christine. Anybody but Jim and Microsoft.
CB: Okay, Brian.
BC: Yeah.
CB: With directory services becoming a key part of the definition of an enterprise NOS, why did Sun flip-flop on their NDS support and scrap that and go with your own LDAP directory server?
BC: I don't know if we ever flip-flopped. We've always viewed the NDS as a great addition to the source environment that you can add on top. We've worked with Novell to do this. We've never really stated that as our strategic direction. What we have done is move from our NIS Plus directory that we've had for about ten years, and we're migrating that over to an LDAP base directory servers. We like the LDAP because it's open. You can get it, you can see it. All the scheme and so forth, they're open for the world to see as opposed to other alternatives up there.
CB: But what are you doing to make that LDAP basic database usable in an enterprise environment?
BC: I'm not exactly sure. Obviously on top of the LDAP directory you're going to have all sorts of management tools that allow you to administer or manage the database scheme and so forth. So that 's all part of the base product with LDAP. And we've had quite a few of these at the NIS World previously. And the trick is that keep the NIS world intact while allowing them to move over to a more modern LDAP-based environment.
GALLANT: Okay, we've picked off the people at the end. Rick, let's work the middle of the podiums here.
RN: Okay. Tamar, I was trying to think about having your job for a day or two. I'll have to admit I probably wouldn't be as witty. And I asked myself here, you know, what can SCO do to better take advantage of the Linux momentum to increase sales of your products. And I got into the question of, you know, you've already provided binary compatibility. So I said, well, why don't you go the whole distance and start providing a user interface like Gnome or KDE. So why not do the whole thing? Why not do that?
TN: So why not provide open source stuff in our products? Why not--
RN: Why not go to providing the user interface and going after the end user market as well?
TN: I mean, I know you know this. But everyone else out there should know that we do actually ship a lot of open source in every copy of UnixWare 7, an open server that we ship. There's a special Commission on Dietetic Registration we have [inaudible]. So that has already long ago [inaudible] some sort of version of [KEE]. Now [inaudible] why don't we have it at--
... browser, you can run it. Why would you want anything else? That's what we ship.
GALLANT: Rick?
RN: No.
GALLANT: Okay. Christine, even for Drew or Eric.
CB: Okay. Eric, all of your competitors are grappling with the issue of how to do service-side management of desktop systems. When it is not stellar in this area, what initiatives are you taking to bring that into your offerings?
ET: [inaudible] four out of five panelists agree, I think, that LDAP is going to be the solution for doing server site management. I mean, we're really lucky to have a really strong partnership with both Novell and Netscape right now. They're both investors in Red Hat. And you get to really pick the brains of a lot of really smart people at those companies.
And I think over the next six months and the next twelve months you're going to see Red Hat Linux, at least the Red Hat version of Linux on the service site be administrable through LDAP. And a lot of the servers are going to start reading their configuration from LDAP in the same way that the Netscape servers have done for about two years. And at the same time we're going to start building more advanced graphical and Web-based tools for doing administration of your LDAP data so that machines can--[inaudible] LDAP hierarchies, you can take advantage of LDAP replication to make it very easy and inexpensive to install duplicate servers and to replicate servers across your enterprise.
CB: And you've got those vendors agreeing to do open source?
ET: No. I don't think open source presented that much of an issue. There is an open LDAP server right now. Maybe it's not the best one in the world. But if somebody has to run their LDAP server on the Netscape LDAP server on Red Hat Linux using Oracle as their back end, that's not an issue for us. The important thing is the LDAP is an open API. It's an open [inaudible] API. It's an open protocol. The client libraries are all open source from netscape today. So we can go with full open source LDAP client and people can use whatever server they like. They can like use the open source server, they can use a commercial server and they get - we're delivering the choice back to the person who is actually implementing this solution.
GALLANT: Okay, very good. Drew, we haven't forgotten you. Rick?
RN: Okay. I'll wrap this up with a fairly straightforward question. Over the last four or five months Novell has really been pushing NDS; NDS Version 8, the caching products. [inaudible] user in a medium-sized company has been a loyal net area user for the last six or seven years. And I'm hearing all this stuff now about NDS Version 8 running on all these operating systems, and caching and Novell going after, you know, being a special purpose product. How am I going to feel about NetWare, and I've been building and running my business on that? What should I be worrying about here in terms of you keeping NetWare alive?
DM: I think you mistake the issue of going cross-platform with NDS with any diminishing of NetWare. It's true that we are putting NDS cross-platform. We're also in an aggressive way embracing LDAP. If you look at our history we typically were there first with networking DOS 1.0 and other things. When people came out with new APIs that were industry standard, we already had the functionality and we just mapped it back to the new API's and the new protocols. And that's continuing today with LDAP. And we'll see how that plays out with active directory. So when we're there with the functionality we can kind of come back the other direction.
Now, in specific response to your question, NetWare continues to be a very strong platform for this class of applications; for caching, and we continue with file and print. And I just view the role of NetWare expanding, continuing to have the same value around file and print. But now add security, add access security, things around BorderManager, the caching and the other things. Those are all very synergistic with your platform. And they're new solutions that our existing customers can really [inaudible].
And if you've noticed the sales of Border Manager and the sales of NetWare 5, it's confirming that that's happening with our customers.
GALLANT: Okay, [inaudible]
GALLANT: Quick follow-up. Now let's change it to an application developer. Why should an application developer [inaudible] be feel comfortable writing to NetWare, when it has those magic three letters "NLM" associated with it.
GALLANT: You know that was going to come up, Drew.
DM: Of course, yeah. If you're writing a particular class of application that's very performance critical, needs to be close to the middle, or if it's very tied to the existing services and needs to be co-resident on the server, then NLM is a good platform. And we have a number of people doing that. If it's less performance critical then Java is fine.
Now, if you want to go - you know, we're not positioning ourselves as a general purpose or less. And so integrate very well with all these other environments. In a sense NetWare is defined a lot more by protocol than we are by API. And so we will - and being defined as such we can exist as we support the open API's or open protocols within these other environments and provide our value that way. So we're not fighting that battle.
GALLANT: Okay. To wrap up our experts panel, each of the experts gets to ask one question of whoever they want. No follow-up; one minute answer. Christine, who do you want to take on?
CB: Jim, where is the exact cost savings and time savings for a network administrator using [inaudible] directory services, reports of thirty some odd different tools you have to use to manage it. I just want to know how is this going to save admin's time and money?
JE: There's one tool. It's called Microsoft Management Console. There are different snap-ins for that console. But in terms of where the savings will be, it will be in terms of good policies and being able to set up things like quality of service on your network; being able to set up security in the hierarchical fashion and using good policies; and in particular being able to integrate that into the complete environment; not just the file server environment, but the Web environment, the applications environment, the entire infrastructure. Because we are a general purpose operating system and we do provide the capability for people to run everything from their network to applications on a single server. Administrators will have reduced costs by administering that complete environment in active directory.
GALLANT: Okay, Rick, who is your target?
RN: Actually I'm going to ask Jim, too. This more of a straightforward - I have an easy question. You were telling me who actually came up with the name Intellimirror. But to go into the real question of I'm company now and I want to have [inaudible], what do I actually have to have to get any benefit out of it?
JE: In terms of the network, in terms of the systems, in terms of the configuration?
GALLANT: I think we missed part of the question because you were [inaudible]
RN: I'm sorry. Intellimirror where are we getting the return on what Microsoft has talked about with Intellimirror? What do I have to have in terms of servers, desktops, network connections--
JE: In terms of Intellimirror it is aimed at Windows 2000 servers and Windows 2000 professional clients. At this point Intellimirror doesn't provide a lot of benefit to down-level clients. I think that's part of your question.
In terms of the exact amount of bandwidth and, you know, exact server configuration, boy, that's an impossible question to answer unless I know a lot about your environment and how much you'll be passing across the wire, and those kinds of things.
What I do know is that it is a lot cheaper to add more bandwidth, build more servers than to send everybody around to put software on desks and generally to manage those desktops. So many customers, I think, will find that valuable.
GALLANT: All right, Jim. Take a breath because you're going to start the vendor questioning. Thanks to the experts. Great questions there. Rick, nice job of filling in. Jim, your whole left side is open. Fire away.
JE: Okay. I think I'm going to start with Brian. Brian, can you tell me a little bit about what happened at D-Day in eTrade recently. You know, you have made a big deal about reliability. And yet some of your most prominent Web sites have had [inaudible] reliability this year.
BC: I think this is going to be an interesting issue for everybody out there in the Internet industry, is what we're seeing is the amount of data being created, the amount of transactions created, is unbelievable compared to what we've ever seen before.
Now, what this says is that even with Solaris we have more to do. What you're going to see from Solaris as we go to the future is kernel level clustering built right in to deliver [inaudible] levels of availability. You will not see that-you will not see that from Microsoft [inaudible] who knows when.
The other interesting point is when you look at this kind of large-scale computing environment, you well may have a tough time taking something that was really built for a workgroup and trying to scale to that level. It's fundamentally architected in a different space, and that's something you have to keep in mind as you look to the future.
GALLANT: You know, Jim, I just bet you have a follow-up to this.
JE: You bet I do. [Laughter.] A couple of things. One is that 60% of SAP customers run on NT today, and that's not workgroup software. Secondly, we do provide--
GALLANT: Is this a speech or a question, Jim?
GALLANT: I guess my response in terms of the question is I didn't really hear the answer. I mean, did you guys have [faults] at those that caused those sites to go down. What are you doing to fix them?
GALLANT: Specifically, I don't know what we have, what specifically happened at eBay. But what I do know is with Solaris what you're going to get is around-the-clock 24-hour-a-day, 365-a-year rapid response team, a level of service that you're not going to get out of a PC company essentially.
GALLANT: It's times like these that the judge wishes he could ask a follow-up question.
[Laughter.]
GALLANT: And he may take that opportunity. Drew, you may ask anyone but Brian and yourself a question.
DM: Okay. Jim. [Laughter.] We hear rumors about how big Windows 2000 is; the 40,000,000 lines of new code. You also hear rumors that some of the key architects have left, giving you a high stock price, et cetera. And you're biting off a lot; a directory a new OS ada-ta-da-ta-da-ta-da. When will Windows 2000 really be solid enough to deploy in the real mission critical? How many service packs-what's the--[laughter]. [Applause.]
JE: The answer is when customers tell us it's ready. We are going to deploy this Windows 2000 in production sites before we ship the software. So customers will tell us when it is ready, and that's when we will ship; the first version, not the service packs and so forth. So we are doing everything we can to make sure that it is a very reliable and very stable operating systems. Customers will tell us.
GALLANT: Drew, follow-up?
DM: Okay. Any idea when that really is?
[Laughter.]
JE: The current plan calls for that to happen by the end of this calendar year. But, like I say, we are not so much committed to a date as we are committed to a quality level, and we will do the right thing by the customers.
GALLANT: Okay. Eric. You may ask a question of Drew or Tamar at this point.
ET: [inaudible] opportunity on the desktop and how you think [inaudible] going to influence your opportunity in the desktop space? You've been saying you're a server operating system. But is it time for Unix to move off the server?
TN: It very well may be. But that's actually not our business. We are a server business, we are a mission critical focus business. There are people, of course, that do put SCS servers on their desktop as workstation environments. That's fine with us. We, of course, encourage that. We have things like AT&T support to enable that. But in fact our focus is on the things that make servers [inaudible] on the reliability, on the scalability I'm getting the data center support in. That's our focus.
ET: Does Linux give [inaudible] more opportunity on the server, as opposed to Windows on the desktop?
TN: Well, I think in some sense, yes, actually. I think Linux on the desktop is special because you have the compatibility for Linux apps. And our operating system allows people who have time on their hands to play with Linux and develop applications and then deploy them on our servers when they're ready.
[Laughter.]
GALLANT: You know, Eric, I was going to compliment you on one of the nicer questions you had.
[Laughter.]
ET: I'm just glad [inaudible] applications for SCO doesn't take any time.
[Laughter.]
GALLANT: All right, let's settle down. Tamar, you may ask Eric, which may be perfect timing, your question, or Drew.
TN: Okay. Eric, I read on your Web site that you know Y2K warranty. And I was wondering if you recommend to your customers that they sell their businesses before December?
[Laughter.]
ET: You know, I have yet to see a warranty from any company for Y2K. [inaudible] IBM [inaudible] is not a warranty. We do have a statement on the Web site saying that we've tested it, we've had lot of people test it. As far as we know everything is compliant. Our statement is almost equivalent to IBM's statement and others like that. I don't think anybody out there is saying, "We're going to pay the cost of your business going down if something happens on January 1st." And that's what warranty is.
TN: Okay. So I may understand, you're saying that you actually are offering the same level of Y2K guarantees that IBM, SCO and others are offering?
ET: Our statement is very similar [inaudible] ones on IBM's Web site.
GALLANT: You've worded your follow-up already. Careful wording is critical at these times. [Laughter.] Brian, we're going to give you a shot at Drew here. But then I'm going to take the questioning this way so that you have an open field coming back. But let's ask a question of Drew right now.
BC: All right. Let's see, here. I think you mentioned earlier about Java and some of the enterprise Java that's going on over there. And what I'd like you to comment on - or I guess it's a question and it's like Jeopardy. We've got to figure out how to put this in the form of a question. But anyway, as far as Java goes how important is compatibility across all the platforms and why is that good for customers? And, conversely, when people introduce incompatible versions of Java, why is that bad?
[Laughter.]
DM: Of course, Brian, it's very important. [Laughter.] I'm a licensee of your company. You know, we've got Eric Schmidt, who just do Java as our CEO. And we just recently announced support for the IBM Web Sphere, which is compliant as an execution environment for applications. So we view it as very important. You know, our struggle. Because of the network platform we run some parts of Java faster. We're continuing to struggle to get other things faster, and we're working with other partners like Intel, as well as continue to work with yourself. And we view that as very strategic for us and we are now releasing commercial products. Our cache appliance, for example, has all the management under Java. We're seeing a number of functionality that it's better implemented on Java than as NLM's. We're using it and enjoying and having fun debugging it with your guys.
[Laughter.]
GALLANT: It was kind of a warm answer. Did you want to do a follow-up?
GALLANT: I'd really like to get just your viewpoint on how customers are hurt when bad-you know, essentially incompatible versions of Java are out there.
GALLANT: Can I make one comment to the panelists. When you're answering the question, if you could speak into the microphone and resist that urge to stand on you rival. It makes it a lot easier to hear.
DM: [inaudible] Java at Community 1 [inaudible] That, unfortunately, happened at some levels with Unix. There's always this catch-up and continued stuff there. Certainly, you know, it's hard to see what Microsoft's [inaudible] is going to be in that regard. You get all sorts of rumors if you believe the magazines. It's hard to call.
GALLANT: You can believe Network World, Drew.
DM: Yeah. But we'll just see. I think the core Java [inaudible], the franchise with IBM, Novell; you know, other people, Sun, pushing it. There's plenty of strength to keep it standardized, and we're certainly committed to supporting Sun in that area. So we'll just see.
GALLANT: Okay. Brian, again, we're taking it this way, and the field is open to you.
BC: Okay. So let me think about this. Jim--
[Laughter.]
BC: In the network world interoperabilities is unbelievably important to every customer out here. And what I'd like to understand is whether Microsoft would be willing to license at a commercially viable price to all the vendors up here, sitting up here, the source code to the interoperability sections of Windows 2000; namely the networking stack, active directory and so forth. So that we together can offer a much more compelling solution in general for the customers out there.
JE: Brian, I think that your point that interoperability is veery important to customers is exactly right, and we've taken that seriously for a long period of time. You know, we certainly were one of the first vendors to take TC/IP very seriously, and we've always been very committed to that.
[Laughter.]
We have also been very careful to provide support for emerging standards, like [Coberos] and other standards in Windows 2000. So if customers need interoperability in Windows 2000 we'd absolutely listen to anything that came from a customer to be able to provide that. That doesn't always mean we need to license our source code. We believe in intellectual property, and we'll continue to take that approach.
GALLANT: Did he answer your question, Brian?
BC: No. So I guess it's no.
GALLANT: Is that your question, "I guess"--
BC: No, no. I just [inaudible] with the answer to the question.
JE: What I'm trying to get out is I'm not sure why licensing the code is necessary for interoperability. There have been standards put together for that interoperability. We're committed to supporting those standards. And that would provide-that is the way we believe that we'll be able to provide customers with he answer that they want.
GALLANT: Okay.
GALLANT: Your follow-up was the answer is no. Tamar?
TN: I have a question for Brian, which is I'm a little bit confused about your strategy. If you are serious about Intel, then you put yourselves out of business. But how can you afford to be in denial of the fact that Intel has won the chip war, and how can you afford to stay in the SPARC business?
BC: Sun strategy is really clear here, is that what we wan to do is provide Solaris on both SPARC and the Intel platform. We do that today in I-32. We'll have exactly the same source code base going forward to I-64. We're the only company out there that has a clear road map to I-64 from I-32 today without any transitions or interruptions.
Now, as far as from a chip standpoint our belief is when you look at the SPARC product line, we offer a very unique set of capabilities with that product line that you can't get anywhere else; specifically not with the Intel platform. Now, it's different strokes for different folks. Some people are going to want to have the capabilities that SPARC base systems provide. Other people are going to want to have what the Intel base systems provide. And as a result we'll sell Solaris to both sides.
TN: I just have one follow-up question that would be - I mean, I would answer it the same way if I were you, except that Solaris and Intel is basically invisible on the market. So exactly how [inaudible]
BC: So essentially what we have today is actually a very rapidly growing basic. We've shipped over 100,000 units to the Intel market over the last six months. So it's growing very rapidly from a very base to a very vast, large and-vastly growing and large install base. So it 's actually succeeding very well. And I'd also put down the gauntlet in that, when I-64 shows up we'll be right there at the party with a fully optimized version of middle price class Solaris ready to roll, ready to rock and roll right there. So I think we're going to be in great shape as we go forward on the Intel platform.
GALLANT: Okay. Eric, you may ask Drew or Tamar a question.
ET: I just want to say being in the middle is not fair.
[Laughter.]
Drew, I would like to ask you - I know [inaudible] talked a lot about [inaudible] We've talked about [inaudible] talk about SMP and they talk about NDS and LDAP. What [inaudible] strategy [inaudible]
DM: Well, we're basing most of our network management around our directory. We've got a very successful product [inaudible] provides desktop management. We continue - that's merging with the ManageWise capability, which was more of a server level device. This is actually a workstation management. It's very powerful with directory enabled. You can basically click one button and potentially install apps or remove apps from many machines in a very powerful distributed way. And it really illustrates the power of directory. So our fundamental focus is around directory enabled management things. Yeah, we've got a bunch of acronyms because we're a large company like Microsoft, and we have a number of solutions, legacy and new, that we struggle with. But you see a consolidation. You're seeing an increased amount of the main Web enabled as well.
ET: Well, how do you [inaudible] talk to the desktop management task force to have this [inaudible] trying to work around the directory [inaudible] despite a directory almost? I'm just curious [inaudible] going to enter into your product line for the next years?
DM: Yes, they are. And with challenges.
GALLANT: Could you be a little more specific about that, Drew?
DM: No.
[Laughter.]
GALLANT: Because we'd be happy to deliver that message [inaudible]
[Laughter.]
GALLANT: All right, Drew, you're actually up, and you ask a question of Eric or Tamar.
DM: Tamar, SCO's foothold, you know, is kind of the low end Intel server line. And it seems like you are [insult] in various ways; by Linux and certainly by NT. What's your real strategy to keep those products, some of the volume there from eroding you?
TN: There's two parts to that strategy. First of all, there's a lot of growth that we have in a small and medium business market that-business that Drew referred to as our traditional stronghold. Not only with our current products growing there, including Unix 4.7, but also moving into new businesses [applying] and serving, which I alluded to earlier, and [value] devices. Those are all typical small and medium business spaces.
However, there's been a very big [inaudible] over the last three years to move the enterprise and data center, the first step of which, of course, is the acquisition of Unix 4.7 from Novell, and the acquisition of me at the same time, and then with the release in March of our data center edition of Unix 4.7.
And at the moment really on Intel there are no other operating systems that can touch us in that space. I would say we're two or three years ahead of the competition here on this [inaudible]. I have no doubt that our competitors are smart and have a lot of money, and I'll be sure they'll try and catch up with us. But by that point we'll hopefully have leapfrogged them.
GALLANT: Do you have a follow-up?
DM: No.
GALLANT: Okay. Jim, meet Eric.
JE: Eric, there's been a lot of discussion about Linux's scalability. When will Linux provide concrete proof in the form of [audited] benchmarks like the TPC benchmarks, or like Spec Web or other audited benchmarks?
ET: Very simply, all the criticism of the Linux 2.0 kernel is correct; it wasn't very scalable. We're working right now with our partners on the Linux 2.2 kernel. We're seeing [inaudible] linear scalability to eight processors until - eight processors on Intel. And we work with Sun actually under UltraSPARC line. We have seen near linear scalability up to 14 processors. And I think as the months go on and we really solidify some of the relationships with our bigger partners, you're going to see the numbers start to appear. And I think you'll be very impressed with what you see. Or disappointed.
[Laughter and applause.]
GALLANT: Jim, are you stunned by that frankness?
[Laughter.]
GALLANT: Eric, could you answer the question: When will you publish benchmarks that prove eight-way scalability which you just claim--
ET: Again, it will be within the next couple of months, and certainly by the third quarter, I think.
GALLANT: Okay. I'm going to go through a quick lightning rod here and let the vendors ask whoever they want a question; see if they're ganging up or if they're spread evenly, and then questions from the audience.
But there is one note about questions from the audience. When you get up I will ask you the name of your company, and this is to prevent sandbagging, which we've had a fair amount of in the past, where people from one of the vendors gets up and ask he vendor yet another panelist type question that they have prepared for us. So you can still ask the question. We just want to know which company you're from.
Eric, can we start with you? We're going to go with one-minute answers here. And I think I'll allow like a 15- or 20-second follow-up if you need it. So ask away to anyone you like.
ET: Jim, IDC says that Linux has about 17 million users now, and Microsoft to support the MacOS platform for a long time with things like Office. And I was just curious, how do you evaluate what operating systems to port to, and when is the Linux market big enough to port Office?
JE: Well, there are two things that we look for. One is actual usage in corporations. And we ask our customers all the time whether they are today using more than 10% of their desktops on Linux. And the answer we get back consistently is no. The second reason, and the second thing we look for is that we need a stable platform to develop on; in particular a stable graphical user interface. And today with Linux it still really hasn't coalesced around one graphical user interface. And so it would still be difficult for us to develop office for that platform.
ET: Well, it's interesting. Actually this has been [inaudible] for longer than Microsoft Windows. Many people consider that a stable API. So I'm not sure I understand that point. But [inaudible] participate like [inaudible] customers who are interested in Microsoft Office, would that help convince you?
JE: I would be happy to accept a list of customers who are actually planning to deploy that. But I will balance that out with my own research into the customer base.
GALLANT: You probably want to be careful with that customer list sharing anyway.
[Laughter.]
GALLANT: Tamar, anyone you like.
TN: Okay, I guess we have to gang up.
...
GALLANT: Well, let me be clear about that. In general we try not to break applications. And as customers moved from Windows 95 to Windows 98, for example, we didn't break applications. As customers moved from NT 3.5 to NT 4 we didn't break applications. Certainly in Windows 2000 we have some challenges there. And we had to make some hard decisions around reliability versus application compatibility. And specifically we see a lot of applications that do things that cause DLL hell. They [inaudible] DLL's in the system directory and things like that. We [prevent] that in Windows 2000, and that breaks the application. But we think we made the right tradeoff there.
TN: I get a follow-up?
GALLANT: You do.
TN: Okay. So you [inaudible] for a liability and yet, as Drew point out, you've added about 35 million lines of new code. How can that be reliability strategy?
JE: I haven't figured that out.
[Laughter.]
It isn't about the number of lines of code. Let me take, for example, a product that I know intimately. With SQL Server 7 we added a lot of lines of new code to that product. But we also tested it well. We tested it and put it in production with customers. And the reaction from both the marketplace and the analysts is that that is a very reliable product. We're going to follow that same best practice with Windows 2000 server. And, you know, we'll see what happens.
GALLANT: Okay. Drew, anyone you like.
DM: Let me ask Brian--
GALLANT: You're sure you don't want to go after Jim again?
[Laughter.]
Microsoft's our friends.
[Laughter.]
GALLANT: Let's get a question.
DM: I'm just wondering how you can afford to fight three major battles. You've got the Chip battle with the SPARC; you've got Unix battle; you're fighting the Java battle and we're with you on the Java battle. But when push comes to shove and volume comes to play and, you know, the SPARC volume versus the other, the server volume versus this, you know, something is going to give at some point. Do you know what that's going to be? Can you guess what that would be?
BC: I don't think anything is going to give. Because I think strategy is a great strategy. And the reason why is on all those fronts you're talking about Sun has taken a very progressive approach on how to do this. So, for instance, on the chip side we actually don't do the fabrication, which is the most expensive part of creating a chip. We do the design and we work with other companies to actually fabricate the chip. So from a-the ability investment ROI on that investment, it's extremely high. So it allows us to be able to design those chips, get them out there without having to make the big investments. That's the first side.
As well is when we look at our fundamental strategy, as I alluded before, is we think that API's and ABI's and so forth [inaudible] industry protocol standards are there for other people also to implement [inaudible]. In a lot of ways we agree with the Linux model. We think that we should be pushing standards out there for other people to also innovate. You know, Bill Joy, one of the founders of Sun's, famous line was, "Innovation happens elsewhere." So the whole idea is to get a whole lot of other people in the ecosystem, with people working around helping to try to make the products better.
So this model is work, and if you look at Sun's financial growth, the product line growth-we grew over 50% last year, year to year, on servers-it's phenomenal. So it seems to be working.
GALLANT: Okay. Drew, follow-up?
DM: No.
GALLANT: Okay. Jim, you're next and then we'll follow up with Brian, the last vendor question.
JE: Okay. Brian, you have made claims for linear scalability at Solaris. And I don't know whether you're saying SPARC or Intel. And yet the facts don't support that. Today the best four-way numbers are about 24,000 transactions per minute on the TPMC and your best number is about 115,000. However, that's only about four and a half times the scalability on a 64-way machine; you'd expect 16 times. And it's about 20 times the price. So how do you prove that you do have linear scalability?
BC: I have no idea where these statistics come from at all; what systems are running or whatever. But what I can tell you is that when you add a processor to-for instance, an Enterprise 10,000, a 64-way system-each time you add a processor you get about 85 to 90% of the new processor's performance. That's much cheaper than adding a whole new system to try to get that performance.
In addition, what ends up happening, which is as you add that up it goes up to about 34 to 40 processors before it starts dropping off to about 85% of each new processor as you add. Which, when you look at it from a theoretical standpoint, is about the maximum you're going to get all the way up to 64.
Now, what I should also mention is you'll see that scalability going up to well over 100 processors next year, without, by the way, the use of clustering. When you add clustering into this, we're going to be 16-way times 100 processors each; all the way up to that level, which is, I don't know. You calculate it. I guess it's 1,600, one thousand six hundred CPU's in a cluster. So this is all going to be happening next year. By the way, this all starts [inaudible] system.
GALLANT: Okay. Brian, you have the last question.
JE: Do I have a follow-up?
GALLANT: Oh, I'm sorry, a follow-up, yes. I'm sorry.
JE: So, Brian, when are you going to publish numbers that prove what you've just said? So you are going to publish numbers on four ways, eight ways, sixteen, thirty-two, so we can actually see the degree of scalability that you claim. Go check them out on the Web site today. We publish all the way up to 64 [inaudible] 10,000. You know, they're there for the world to see.
GALLANT: Okay, Brian, anyone you like.
BC: Jim--
[Laughter.]
So there's a fundamental premise in software development that Fred Brooks developed when he was at IBM with Professor Brooks, called - made famous in the book The Mythical Man Month - a great book, by the way, for everyone to read. Basically what it says is as you increase the code base of a product it's extremely difficult to make reliable at that level. So out of that basically there's a fundamental principle; namely, you either increase reliability and keep features under control, or you risk reliability and add features. Those are the two options you have in software development.
The question is, what has Microsoft invented that is going to allow them to break that rule with Windows 2000?
[Laughter.]
JE: I think you misinterpreted The Mythical Man Month, which is much more about adding people than anything about as you described it. What we essentially do is try to build small teams of people that are focused on very particular parts of the code. And we try to keep those teams small because we do realize that you can get too many people and it can get out of control.
In terms of insuring reliability and [inaudible] customer' needs for features, which customers do need, we do that essentially by hard core testing both internally and externally with customers.
One other thing I might mention is that I do agree with your point that as you increase the number of permutations it does get harder. So, for example, one of the things that we focused on with our data center product is limiting the number of configurations that it will run on so that we can insure very very good reliability on those configurations.
GALLANT: Follow-up, Brian?
BC: Yeah. As far as the longest line in the software industry, we've all seen projects from hell. OS 360 is one for IBM, Copeland was one for Apple. I should dare say Solaris 2 in the beginning for Sun was one like that. And my question is do you think that Windows 2000 should be donned with the official label "Software Project From Hell"?
JE: Well, it's interesting some of the ones that you mention, like OS 360, for example, was a project that delivered for IBM many many years of customer satisfaction and a great product for customers. So in the sense that it's like OS 360 I'll grant you that.
GALLANT: Did he just say it was the [inaudible] hell?
[Laughter.]
GALLANT: I think he said it's like OS 360.
GALLANT: Okay.
[Laughter.]
All right. Let's take some questions from the audience now. If you feel like you can really yell it out, you can stand up. Otherwise, come to the microphone. Who has questions. Right here. You want to stand up and-who are you with?
Q: [inaudible].
GALLANT: Okay.
Q: Basically I love what Red Hat Unix opening code. Really helps out programmers and administrators. I think that what Sun has done with the Java [inaudible] write down those lines. My question is, is Microsoft going to learn [inaudible] Are they going to implement the Java API or are they going to use it as a tool to get people to go to [inaudible]
GALLANT: I think that's for you, Jim.
JE: Okay. There's a number of different questions there. But we do implement today Java virtual machines in our operating systems. And we will continue to do that.
Q: [inaudible]
JE: No. We do today. Java virtual machines exist in our products. And it is one of the languages, for example, that we'll support. But you mention DB. We don't think that customers are going to retrain all of their developers in Java. We actually think that you have developers out there who know DB, who COBOL, who know C and who know Java. And so our strategy is to have multi-language support in our product including a Java virtual machine.
GALLANT: Did you want to ask a quick follow-up to that?
Q: You're not using [inaudible]; you use your own. And it doesn't work on Sun's.
GALLANT: Is there a question in that, though?
Q: [inaudible]
JE: Well, first of all, there's a mistake in that, in that the courts require that we have exactly Sun's controlled API. Nobody else can innovate; just Sun. They're the guys who own, you know, everything there. We can possibly do any innovations. So that's what is shipping today in the products; just any ideas that come out of Sun.
GALLANT: I think you touched a nerve there.
[Laughter.]
GALLANT: This gentleman right here. Go ahead. What company are you with, by the way?
Q: I work for the Children's Home Society in California.
GALLANT: Okay. Why don't you step to the mike. It's right up there; it will be easier. And remember, to a specific vendor, please.
Q: Okay. This one is to Jim. In our environment we deployed a lot of empty Workstations 4.0 in preparation for networking. What I'd like to know is in Windows 2000 how many of the new features of Windows 2000 can my NT 4.0 workstations use without having to upgrade those workstations to take advantage of it?
JE: Okay, so the question is about if I put Windows 2000 servers how many of the features of NT 4 clients [inaudible]--
Q: Yes, can I take advantage of?
JE: For the most part the only key features that won't be supported there are the Intellimirror features. But active directory, for example, will be supported; the improved networking will be supported; quality of service to the desktop will be supported. So the only major feature where we simply weren't able to do it in the amount of time that we had to get it back to NTW 4 is the Intellimirror technology.
GALLANT: Do you want a follow-up? Okay, right there in the blue. What company are you with?
Q: [inaudible].
GALLANT: Okay. And who is your question to?
Q: Jim.
GALLANT: All right.
Q: [inaudible] What [inaudible] does Microsoft have that are different from the NT 4 roll-out that exist [inaudible]
JE: Yeah, it's a tough problem. The first thing we're going to do is we are going to provide a lot more training both to our customers and to the channel. We announced a couple of weeks ago that we are investing $40,000,000 to subsidize training to our channel and our customers. And that is all about how to roll out the product. That's not usage training; it's all directed to systems administrators and IT professionals, to help them understand that.
The second thing that we're doing is that we are publishing--[inaudible] calls them "Redbooks." They're really configuration guides and deployment guides. The first one of those will be published at our tech American Dietetic Association conference, which is coming up in two weeks, I think. And then others will be published as we come up to deployment. Those are all based on the real world experiences that we have with RDP customers. So we think it will apply to the right tools.
GALLANT: Did you have a follow-up? Okay. I think we have time for two more quick questions if we could. Right here. And you are with?
Q: [inaudible]
GALLANT: Actually you need to target that to a specific vendor. Otherwise it will take too much time.
Q: Okay. Novell. [inaudible]
GALLANT: Well, for sharing data between Macintoshes and PC's we were the ones who really defined that market, and we continue to provide good support through some partners in terms of supporting sharing of files and data between the two environments. Certainly as things go more Web and da-da-da-da base, in some sense it doesn't matter what the server is. But historically and traditionally our focus has been around networking all products from all sorts of vendors. And so our heart is there more than I think anyone else.
GALLANT: So follow-up, you might hit another one of those vendors.
Q: [inaudible]
GALLANT: We are very committed to supporting a wide range of clients, including Macs. We supported Macs networking for, I think, about five years; four or five years now. We are also committed to supporting you in terms of applications. And we've been trying to do a good job of getting our applications for Windows and for Mac out on a more similar time frame. So we think we're doing the right thing by the Mac community. If you have feedback that we're not doing some of the right things I'd certainly be willing to listen to that.
GALLANT: All right. One more question from the audience. I saw you first. You go right ahead. Your microphone is right there so you might as well use it. And, I'm sorry. Who are you with?
Q: [inaudible] Industry.
GALLANT: Okay. And your question is to?
Q: Novell.
GALLANT: Novell.
Q: With the current releases of your NetWare 4.2 and 5.0 you removed support for Macintosh clients and made Macintosh clients purchase the software in order to use your server. And I still believe today that the software is not available to even purchase for the server site for Macintosh. Would you care to address this issue?
GALLANT: That was an unfortunate decision that I personally would probably challenge within the company. But we felt, given the time frame and stuff, that it was best for a third party, who is a very strong third party, to do that. And quite frankly I would hope that will get reversed at some point.
GALLANT: You all set?
Q: Yeah. I guess the follow-up would be when will that be reversed?
GALLANT: When will it be reversed?
GALLANT: Depends on how hard I can argue. Novell had a challenge a few years ago, and we're just emerging out of that and focusing. And the Macintosh-certainly with the resurgence of the new Macs and stuff it demands another discussion. I haven't talked to Steve [inaudible] in a couple of years, and it sounds like a good excuse.
GALLANT: Well, Drew, we're behind you on that if you need our support. We have closing statements and then a couple of remarks at the end. So, Rick, what were your thoughts on what you heard today on your rapid fill-in basis?
DM: Well, I did want to thank all of the panelists. I think they were very aggressive in getting their information across and trying to make their points.
The thing that over all I draw out of this-and it's as much from my research and from the responses here-is I think the operating system market is about to go through a sort of a crunch time. And there will be two different trends, the flow-through here. One is scalability is going to be redefined here. Things like Hotmail and various Internet applications out there, bringing these systems out to support millions and potentially billions of users or systems, is going to redefine what all these vendors are doing. And I think that a few years from now not one of these vendors is going to be able to stand up here without having to address a problem in the Internet that was very visible to people. All of them are working hard on it, but this is something where they are dealing as much with an unknown as anybody else.
The second part is I do think you're going to see a move away from general operating systems doing everything. And it's interesting both Novell and Linux and somewhat SCO are going to start being more specialized. And I think you're going to see that trend as well, as people really focusing on addressing a specific part of the network and system needs, and some people really doing well by focusing on sub-sets.
GALLANT: Okay. Christine, what are your thoughts?
CB: I'm going to piggyback on your thoughts of the specialized areas. It seems that Sun is going after the traditional area of the high end scalability database and Web server apps. And SCO is focusing on anything that Unix can do on Intel. Novell is moving NetWare toward being the platform for network appliances. Linux is trying to define itself as the application development platform and then trying to be a player on the desktop against these servers. And Microsoft is trying to do it all.
I guess the question that remains open is how are all these specialized operating systems going to provide interoperability in the future?
GALLANT: All right, very good. I want to take moment to thank some people. First I'd like to thank the audience. It's a great audience, great reaction to the questions, very good questions. I appreciate all the time that you spent and I hope you found it valuable.
I'd also like to thank Cheryl, who is down here in the blue and kept the time and did a lot of setup here and a lot of work there.
Thanks to our panel of experts, Rick and Christine; Rick especially for filling in at the last minute. And Christine had done a lot of preparation work with these questions.
But I think our thanks really have to go to Jim, Drew, Eric, Tamar and Brian. I have to tell you it's a really difficult thing to get up here and field these kinds of questions and to be able to respond in an unformatted environment. I think their willingness to do that says something about their companies, and I think you should go find out more about these products at their booths. But it's a tough thing to do. They did a great job with it and, as I said earlier, they're up here because they're the leading vendors in the industry, and they had to be on the spot for that. So a big round of applause for them.
[Applause.]
GALLANT: That's it. Thank you very much.
[End of Recording]
RELATED LINKS
Our report from the showdown. Network World, 5/17/99.
N+I roundup
More news from the show.
