Surviving Picasso
From falling out and then making up with Microsoft to the ICA and the thin client ... Ed Iacobucci of Citrix tells all.
Just when you thought it was safe to go to the pub and talk about the thin client ... Ed Iacobucci of Citrix tells all. client-server systems to your non-computing mates, it all goes pear shaped.
Typical. One minute the idea looks like it will never go away - the technology is tried and the concept tested and the next minute someone has gone and upturned the apple cart - again. So now we have to learn a new set of initialisations like ICA and RDP, while the term thin client is being bandied about with gay abandon.
Things have become very interesting. When Ed Iacobucci, founder and chairman of Citrix, visited the UK recently to talk about ICA and thin client, the journalists allowed to sit at his knee received a 260-page book from Microsoft Press called Understanding Thin-Client/Server Computing, which on the cover states: 'An overview of hot technology from Citrix Systems in co-operation with Microsoft.' This highlights just how close Microsoft and Citrix are and suggests that some think this new technology isn't going to take a mere two page white paper to explain.
Despite the book, the concept behind the technology is simple. Based on the client-server model, it offers low bandwidth across a network but high performance at what can either be a PowerPC or a dumb terminal. A Windows NT server runs several Windows sessions and each terminal connected to the server participates in one of these sessions. The data transmitted down the line is effectively screen information to the terminal and keyboard information from the terminal back. This means the bandwidth is kept low and, depending on how the server is set up (typically a multi-Pentium system with upwards of 250Mb of memory), the response time is fast too.
Continuing the 'something for nothing' feel of technology, it is worth noting that there is client software for most systems, including Macintosh, Unix and Dos. In fact, it is possible to have a 286 under Dos (if it has a VGA screen to handle the front end) running a full suite of Office 97 applications - a bizarre thought. It will come as no surprise that Microsoft has taken a lot of interest in Citrix and has - after a falling out last year - bought a six per cent stake in the company as well as licensing some of its products.
Citrix, due to its proximity to OS/2 and therefore by association Windows NT (back in the late 1980s NT was originally intended to be an OS/2 release), has managed to play with the innards of NT to turn it into a multi-user operating system.
It's worth noting that in the mid-1980s Ed Iacobucci, an IBM employee, headed up the joint IBM-Microsoft design team that developed OS/2, suggesting he might have some insight into how these things work. And Microsoft is now playing with the technology itself (see box out page 82) and will be getting some of the Citrix technology to us this year. Iacobucci has come to London to spread the word, and perhaps to make sure that Microsoft doesn't steal the day with its own thin client announcements.
Currently, Citrix is doing quite well. According to Iacobucci, it started shipping ICA servers in August 1995 and by April 1997 there were 500,000 licensed users. In September the number grew to one million licensed users, on about 60,000 servers, thought to be about three million terminals.
Thus the first question from PC Dealer has to be: 'what is all about?'
It is about a different way of handling applications. It isn't about computing at your fingertips, but information at your fingertips. The paradigm of the delivery information model is up for grabs, and currently there is no clear solution. There are a lot of things being re-evaluated in terms of how I deliver an application to you and how I make sure my application is available on a 24-hour basis. There is no room for downtime.
It is about predictability, reliability and availability. Does that sound familiar (laughs)?
Well it does sound familiar. Where have we heard it before?
These are things we used to say about large systems - mainframes. It is about stuff we have known about in the industry and seem to have got away from. What we are proposing isn't a product, but a different way of doing things that maybe all the segments of the industry - Microsoft included - aren't going to rush to embrace, but the customers are.
How do you propose to make WinFrame interesting enough to get users away from Microsoft's Hydra implementation?
By creating something bigger that NT 5 plugs into, like the overall thin client system that manages a series of offerings and products that work together, independent products allowing for things like security management, performance management and bandwidth management. The Microsoft implementation doesn't support ICA or the ICA management systems, load balancing, bandwidth management, performance management or applications deployment. It won't have clients for Java, Macintosh, Unix machines, it is simply for NT to Windows communications. Why are they doing that?
OK, why are they doing that?
That's part of the agreement - we licensed the kernel to them if they agreed not to build cross platform compatibility to these other environments.
For example, there is a word processor in every copy of Windows. It is called Wordpad and is very useful, but if you are going to write an article for a magazine you are going to want a word processor. OK, so in this example you buy the word processor from Microsoft, but for the time being we should ignore that. I hate to say this because I don't want to sound like I'm boasting, but WinFrame isn't a product, it is more like the centre of a movement. There is a lot of stuff happening around ICA, WinFrame and some of the related technologies, that transcends the way we do business today.
Picasso is your new product.
What exactly is it?
Picasso is WinFrame 4 Hydra. It probably won't be called that but that's its name at the moment. We licensed NT from Microsoft, made it a multi-user kernel, added all the capabilities around it and that's what WinFrame is. Microsoft said: 'We want to be in that business', and we said: 'We will license back to you the changes we made to the NT system to make it a multi-user platform you can deliver that to the market while we deliver everything else that sits on top of it'.
What happens to Citrix after NT 5 is released?
If we work with our partners correctly, we can expand the role of this beyond solving the business problems and into an enabling capacity, to build a real information appliance.
What is an information appliance?
To me, it is something that connects into a managed complex and delivers information. There is an independence provided by a protocol like ICA, from the device, appliance and application, so you can upgrade the application and servers all day and it can't invalidate the hardware device.
Eventually, we envision the ICA protocol becoming a public standard, but not until it stabilises. The worst thing that could happen is that it becomes a standard prematurely and then goes off in 17 different directions.
But if you send more data down the line, won't that stop it being a thin client?
It is still a thin client. What makes a thin client thin isn't the thickness of the data or the richness of the media, but predictability. A TV is a thin client because when the full programming cycle of the BBC comes out you don't have to take the TV apart to check how much memory it has or run any diagnostics. You know your TV will be able to run the new content.
We want to achieve the same level of predictability for information appliances.
Will this be the demise of the PC?
I would say that nothing will be the demise of the PC. But the general population, or the market of users that need access to information, is far larger than the number of people who want computing. There is a big difference between the delivery of information and applications and the delivery of computing resources.
People are using WinFrame for deploying line of business mission- critical applications. Are they using it to deploy word processors? Some, but is that what you buy WinFrame for? No, it doesn't make any sense.
A good example is Bell Mobility in Canada. Two and a half thousand people across Canada that have activation, customer service and trouble ticket reporting the core applications in the system. They have taken the set of applications and are hosting them on WinFrame servers. They have 120 of them in Montreal and Toronto and people can have their applications on their workstations. But if they want the applications that are delivered by the IT organisation they have to log on to the server farm.
One of the interesting side effects is that they went back and measured it and found the availability of the applications - and remember these are their figures not ours - was merely a fraction less than their own.
BLOOD ON THE WALL, BUT CITRIX SURVIVES
Last year there Microsoft and Citrix had what looked like a falling out.
Apparently, Citrix had made the idea of thin client interesting enough to catch the eye of the Boy-God Gates and make him want to involve himself in its future - was it as bad as it seemed?
We've gone through a number of near death experiences, the first being when we shipped our first product. We called it Multi-User OS/2 and there was going to be a lot of leverage in it. The first day we shipped the product to manufacturing, the Wall Street Journal ran an article entitled 'Microsoft and IBM are getting a divorce'. We spent $50,000 positioning something and it cost $5 million - we didn't have to unposition it. So when this thing with Microsoft happened last year - and I have to say I don't remember saying this, but someone who was with me insists I did - some reporter said "this has got to be the worst thing that ever happened to you, right?" and shoved a microphone in my face. Apparently I said: "Nope it doesn't even make the top 10". Although I don't remember saying it, I believe I would have said it because that's how I felt.
So what caused the fall-out?
One of the agreements we had with Microsoft was that if it was ever to consider building a multi user Windows NT, it had to notify us at least six months before anyone else. In February 1997, it indicated it was contemplating that, so I was forced to disclose that to our investors because it isn't something you keep a secret.
How did you get back with Microsoft?
I took a team to Redmond and we were there for three months. We hired several apartments in a complex and met with Microsoft daily until we worked out the details. We got the deal we wanted and the deal I think is best for Microsoft and Citrix. Microsoft essentially got the inner workings of our modifications to NT.
The teeth of the Hydra - what MS gets out of it
It is likely that Microsoft couldn't believe its luck. There it was, wondering how on earth it was going to manage to get a multi-user angle on its operating systems, when along came Citrix with a big soppy grin on its face boasting how it had made Windows NT into a multi-user operating system.
Citrix had special dispensation from Microsoft to fiddle with the NT kernel code in order to produce WinFrame. Inevitably there was a downside, because it had to modify the actual source code of Windows NT to achieve what it wanted every time. Microsoft upgraded NT and Citrix had to produce another version of WinFrame. So, taken with the result, Microsoft bought into Citrix (six per cent is owned by Microsoft) and signed up some serious software.
It has also promised not to develop its own multi-user software for two and a half years. Microsoft's first version of multi-user NT (which will be built on top of NT 4) will be sold as a standalone product and probably won't be available until the end of 1998. It is likely to be called Windows NT Terminal Server.
Producing a multi-user version of NT 5 is much more involved and will be taken on by Microsoft's own development staff, which arguably doesn't bode well for the finished product.
As well as working with Citrix, Microsoft has also enlisted the help of French systems house Prologue Software. Prologue is the only vendor other than Citrix to have developed a multi-user NT kernel. Apart from help in developing the core operating system modifications necessary to support multiple simultaneous remote sessions, Microsoft has decided to go its own way when developing the client software, network services and protocols necessary for multi-user support.
Instead of adopting Citrix's own popular ICA (Independent Computer Architecture) protocols and client software, Microsoft is creating a new protocol called T.Share, loosely based on the T.120 video conferencing standard being used in its NetMeeting video conferencing software.
Purchasers of the new multi-user versions of NT will automatically receive T.Share software components, to be installed on both NT hosts and Windows clients. The protocol used instead of ICA is RDP (Remote Desktop Protocol).
There are already reports that beta testers think RDP doesn't cut it and is too slow. This doesn't seem bode well either for users that are planning to build their solutions around Windows Terminal Server or for Microsoft.
So, while the likes of Citrix continue to provide ICA, the multi-user bandwagon will continue to trundle and it will be to the benefit of Citrix if Microsoft can't get RDP to work.
All the same, for the simple action of investing in Citrix, Microsoft now has got a serious grip on a problem that was going to cause all sorts of trouble in the near future. After all, what's a hundred million dollars in licensing fees paid to Citrix compared with the headache of developing the thing yourself?