In June, I spent several days on the Microsoft campus talking with Microsoft executives about the impact of realtime and the emerging era of cloud computing. My conversation with Chief Software Architect Ray Ozzie began with a discussion of the recently unveiled Google Wave, now being rolled out for testing by some 100,000 users. Ozzie followed up on his Churchill Club chat, where he described Google as taking on such a hard problem that it might limit adoption:
RAY OZZIE: But what I really meant was that, if they haven’t said that they’re taking on the goal to replace e-mail and IM, then what I said was irrelevant. Like the notion of taking on e-mail and IM means that you have to have a simple protocol, because there’s going to be lots of implementations of them. If that’s not your goal, you can build as complicated system as you want. But if you’re going to do something that is going to be that ubiquitous and that timeless, it’s just got to be a lot more nuggets (of that size ?) —
STEVE GILLMOR: There’s a conflict between them opening it at some point and their stated use case from the beginning.
RAY OZZIE: I mean, when you work through all where they’ve got Google IDs federated, when you look at the UI and how the actual scenarios would actually pan out, the level of complexity on the back-end to get all of that to actually work so it’s easy of the UI, it’s just hard. We barely can get people to use sender ID on e-mail to validate things.
STEVE GILLMOR: Trying to figure out how to be able to go up higher in a conversation and inject yourself, and then see what swarms around that is interesting, but it sort of devalues everything that’s below it. So, there are social cues that you’re sending by doing that kind of thing. I don’t think they have any idea where that’s going to go, and that’s going to take a lot.
RAY OZZIE: Whenever you innovate like that, you don’t know what you don’t know in a lot of dimensions. And like I said, I applaud innovation. I really like that in terms of experimentation. But when you do that, I just know from the Groove experience most recently, from the Notes experience before that, when you create something that people don’t know what it is, when they can’t describe it exactly, and you have to teach them, it’s hard.
STEVE GILLMOR: Collaboration isn’t really an intuitive act for people, is it?
RAY OZZIE: Not really. They teach us — or we teach our kids very, very young to share and to play nice, but the reality is the most effective collaboration systems let people facilitate collaboration even though it’s a human characteristic to want to communicate with people, as long as it’s serving your own purpose.
We did this at Groove We would poll people asking, “what’s the most effective collaboration tool that you actually use?” We expected to get the answer e-mail, because e-mail is the 99th percentile tool that we use for collaboration, but people don’t view it that way. They view e-mail as personal: I’m sending stuff that I want to be sent, I receive stuff that I should see or that people —
STEVE GILLMOR: Right. It’s a pathetic attempt at controlling the universe.
RAY OZZIE: But basically as long as you can couch something — people don’t like to work on things that are joint objectives; they like to work on things that are their key — that satisfy their KPIs, their objectives, not necessarily the joint ones. So, it’s playing into that dynamic.
And all of the systems, as long as I’ve been working in this area, the picture that I’ve always had in my mind is kind of three overlapping circles of technology, social dynamics, and organizational dynamics, or people, organizations and technology. And any two of those is relatively straightforward and understandable. When you mix technology and people, you have something like Facebook or IM. You’re dealing with the dynamics around what is it like to build a tool for people to talk to people. Then you mix in organizational dynamics, and people communicate way differently. Like they don’t post things like they would online because they might get fired. There are dynamics in an org where you say something and somebody will use it politically against you. You might disclose some confidential information. There are different dynamics when you mix in that organization components.
So, even if you build a great tool and it works on the outside, it doesn’t mean that it’s even going to work on the inside of an org.
STEVE GILLMOR: So, what do you think about this real time thing? Do you see it as a thing, as something that has its own —
RAY OZZIE: Is it important, is that what you’re asking, or is there something different —
STEVE GILLMOR: Is it a fundamental thing that’s different, that’s been instantiated by the ubiquity of broadband coupled with the ability of XML based technologies to orchestrate processes that look a lot like social interaction?
RAY OZZIE: I think the answer is yes, it’s important and there are a lot of very interesting things. I think we don’t really know yet which ones are going to be sustainable killer app type usages versus not. It’s really hard to scale things that are at that real time level, and I frankly don’t think we’ve even scratched the surface of what real time means.
When you’re Tweeting only once every, I don’t know, how often do you think the speediest people who Twitter are doing it over the course of their waking hours, if you averaged it out, once every —
STEVE GILLMOR: Well, noisy — Scoble is 100 a day.
RAY OZZIE: Is it 100? Okay. But that’s still not much in the grand scheme of things if you think of how many seconds he’s awake per day. It’s still only once every N seconds.
What if your devices were Tweeting on your behalf to serve you? What if your phone, your car, your — I don’t know your glasses, but different things in your life were posting informational updates that went to services that were acting on your behalf? It’s a perfectly reasonable, realistic thing that could happen if you had an infrastructure that was a message switching infrastructure in real time. It’s a logical direction that things would go.
But my gosh, the infrastructure implications of switching that many input messages through a fabric with arbitrary subscribers, programmatic subscribers and human subscribers to all of those messages, it’s a big computer science problem, really big. And the people who know the most about that right now are the people who write systems for complex event processing for NASDAQ, and people like that. It’s also got similarities to Twitter in that they have to not only have standing queries on streams but they also have to do persistent stored queries to see what’s happened in the past to match that same query. So, they’re just hard systems.
And I’m very optimistic, I think it’s great, I think there are going to be lots of different applications, some in the social realm and other things.
But that’s why I said you have to look at the specific application and what impact it’s going to have. It’s going to impact search, because the nature of search ranking at least classically is based on looking at what’s relevant by what it’s linked to, and if a lot of the links are moving from more static media into more dynamic media —
STEVE GILLMOR: Oh, so you agree with my thesis.
RAY OZZIE: Yeah.
STEVE GILLMOR: Excellent.
RAY OZZIE: I don’t know to what degree the ranking will be impacted in that realm, but I think that’s a fairly important thing. And the Tweet itself is more or less the anchor text of the link and so on.
You know, will it subsume IM? Maybe, because IM is a pattern of communications, a small subset of a pattern of communications that’s kind of like an application of such messaging infrastructure. One person talking to one person could live in that same kind of infrastructure. It’s a subset of many to many or one to many and so on.
I think we’ll ultimately use it, as I said, in the device realm. I think we’ll use it as a notification channel where something happens and we just need to get something out to a subscriber. So, I think decomposing some of the things that we do in our clients where Outlook might pop up a little alert, but it’s also an app, instant messaging might pop up a little alert and it’s also an app, maybe we’ll have alert apps or the OS itself will alert or maybe your devices will have something like that.
STEVE GILLMOR: That sounds a little bit like a widget framework, [where] the apps don’t retain control over these features.
RAY OZZIE: On our current path the apps do retain control, but the real question is, if you have lots of apps that each has some kind of notification, what I learned in the Groove experience is that if you have lots of things that notify things, the user wants to suddenly aggregate them and tune which ones they want high bandwidth instant notification and which ones they only want to occur every once in a while. So, having messages go through something where they can say, okay, I’m going to have all of these different classes of messages, move these classes into this bucket, these into this bucket, notify these on my phone, these somewhere else, it’s useful for the user. So, I think we’ll ultimately go in that direction. We’re not on that path right now.
STEVE GILLMOR: So, your concern about the overwhelming fire hose aspect of this that is just difficult to scale up to that kind of —
RAY OZZIE: Well, there’s a technological aspect and a human aspect. From a technological aspect it’s just a hard computer science problem like some of the Azure things that we’re doing or some of the things that Google has had to cope with in high scale systems. Anybody who has built a really high scale system — Messenger or Yahoo! Messenger — whenever you’re dealing with half a billion users, there are some interesting scale issues. And that’s simple point to point. If you take it end to end, it’s just even more.
But beyond the technological scale issues, the reason I was getting at the unified notifier concept is because I think as humans we have these issues. And certain of the events, certain classes of the events we want to treat, as Dave says, like a river where you don’t really care if you miss something, you know. It’s where you’re not trying to keep up every little thing. It’s maybe it’s an amusement, maybe it’s just a background activity.
Some types of events you just want to see them. You just don’t want to miss even a single one in this big flood of notifications. And so we just need better tools.
STEVE GILLMOR: I always thought it was interesting, at least in the initial design, that Mesh had the contacts feed. The implication of that was that you would be able to apply social ranking to messages or to the messages that those people are interested in, and develop a way of being more efficient about finding stuff that you particular in your affinity group are interested in rather than this sort of broad Digg-like stream of incomprehensible.
RAY OZZIE: Well, in the Mesh case, in the Groove case they’re both tuned for smaller group interactions, and those tend to be ones where you are interested in the types of alerts, the types of things that people do.
STEVE GILLMOR: But isn’t the use case for each individual person more around figuring out what the people that you communicate with and are interested in, what they think is important rather than trying to analyze a volume of information which guaranteed is going to be irrelevant?
RAY OZZIE: I think they’re both important to different people in different mixes. How I’m interested in things that happen in my family or my close knit family is going to be a certain priority. My colleagues will be a certain priority. The larger Microsoft will still be a higher priority than what’s going on in some of the feeds that I might subscribe to or individuals that I might follow in Twitter, and it’s not an absolute of one being more important than the others for every individual.
J Allard describes it as mosaic identity; we each have different facets, and it’s hard to make generalizations about what’s important to one person or another. My son is really into gaming, and so he follows a certain community very, very well. Even though it’s a public thing, he doesn’t want to miss anything. Whereas my public stuff, I’ve just got such overload, I can’t pay attention to what’s going on out there as much as in the circles that I just have to do.
STEVE GILLMOR: What Microsoft is doing and this real time moment that’s occurring with Twitter and Facebook — Do they intersect, and if so, how? This private/public hybrid stream of messages, a mashup of email blogging.
RAY OZZIE: There are certain aspects where I think it’s really adding to the palette of technologies upon which we build things, but that’s kind of orthogonal to the public versus private stuff, that the nature of the tools you build where something is default public or default private. Blogging was at its birth default public, and so that had a certain impact on the vast majority of how people used it. E-mail was born as a default private mechanism. We don’t have many mailboxes inside of Microsoft at least where you can go to somebody’s mailbox and look at everything that people are sending to them, and getting from them, seeing who their followers are and who their friends are. It’s not necessarily directly related to real time; it’s that they built an interesting tool and happened to make those default choices. And it’s really interesting what has happened.
STEVE GILLMOR: FriendFeed, have you looked at that at all?
RAY OZZIE: I’ve only played with it a little bit.
STEVE GILLMOR: It’s kind of like swarming in Groove basically. You see people and instead of there being a little green light, they say something and somebody else responds and you see it, it comes in, into the conversation, and things sort of expand.
RAY OZZIE: Watch it expand, yeah.
STEVE GILLMOR: It’s of a quality that is different than IRC or e-mail or IM.
RAY OZZIE: The swarming characteristic absolutely, you’re right on in the notion that real time does permit a level — peripheral awareness and real time bring swarming, are enablers of swarming. You can’t have swarming without peripheral awareness. But once you’ve got that, then it does —
STEVE GILLMOR: And the implications of that for business and entertainment and a number of different areas seem to intersect with what Microsoft has as resources. And doesn’t there need to be some sort of attention paid to how that can be integrated and not subsumed but rather expressed in what Microsoft is doing?
RAY OZZIE: Yeah, real time infrastructure is an important component of the systems that we build or will build in the realm of communications, collaboration, information, discovery, trend discovery and so on. The work that’s going on in Bing is informing the work that goes on in SharePoint. The work that has happened in one instant messaging system related to awareness informs some of the things that go on elsewhere in the enterprise.
STEVE GILLMOR: Where are we going to see it first?
RAY OZZIE: Within Microsoft? You’ll probably see it in the consumer properties first.
STEVE GILLMOR: You mean in gaming?
RAY OZZIE: It will start in the realm of gaming because gaming tends to be very twitch-level interactive. But consumer communication properties — I mean, the nature of communications nowadays — I’ll give Facebook probably the most credit here. We used to think of communications as relatively partitioned from other modes of things that we do, meaning it’s a message and it’s addressed and a message that you would send to someone. But Facebook, because of their investment in photos, because of the nature of how they implemented the Wall and the little previews of things, and then ultimately apps, they’ve blended social media and communications in a way that kind of informed everyone that communications are richer when you bring other media into it, and I think that’s tremendous.
As an industry all of our communication tools are now getting much more blended in terms of what is media sharing versus what is pure communications. You’re just giving more context to people as you’re communicating with them.
So, you’re going to see more and more innovation probably first in our consumer properties, and then that always bleeds over into our enterprise properties.
How will video and use of video impact how people do things within an enterprise? Yes, from a training perspective, but I’ve, for example, seen lots of use of screencasts as an internal tool. Instead of having a meeting with someone, you just do a quick screencast of some things that are points that are important to you on your screen, narrate it a little, send it over, it’s only three minutes, and you can save a lot of time and get a lot of information communicated to someone. Embedding that in an internal blog, you can communicate lots of things in a very short period of time to many people.
I think one of the fascinating things that’s happened over the last few years is that the tendency toward short messages, not long blog posts or long documents, the snacking on little bits of information is really helping people in the enterprise. Once we’ve cognitively gotten over it, it’s just letting us be aware of more things that are going on in fairly complex situations.
I’m going to really bore you here, but there’s a theory — you know, there are theories that go back to the ’30s with Ron Coase about transaction cost economics, and how the nature of an organization or any organization of companies, whatever, to do a common work product is simply based on you can distill it down to the number of interactions to get a piece of work done, the complexity of the number of communications, interactions that because each one — each delay and each hop represents an opportunity for distortion, like playing telephone —
STEVE GILLMOR: Yeah, it’s the old sermon on the mount joke.
RAY OZZIE: Exactly. So, if you have a communication mechanism or an interaction mechanism, a collaboration mechanism that reduces the distortion or delay in getting common context between the people who have a common work product, it’s of value to the whole group, the whole set of people who want to do that.
Every few years there are certain trends, certain things that get buzz at any given moment in time. I think there is a tendency to look at what’s going on with real time technologies, real time streams of events, and say, is this a flash in the pan, is this a technology that will come and go?
By way of analogy, the excitement around peer-to-peer, people would say, was that going to transform the industry, how is this going to impact consumers, how is this going to impact —
STEVE GILLMOR: Well, you kind of launched Groove on that, on the back of that, even though that wasn’t the point.
RAY OZZIE: We did. We started it several years earlier, but certainly we took advantage of that when that was a spiking trend.
I would say that I have a higher degree of confidence that real time technologies are going to broadly impact systems. It may not be in the form of a Twitter or something like that, because that blends real time and a communication tool, and it is the pioneer and we’re all kind of looking at it trying to understand what’s going to be sustainable and what’s not. But I do believe real time notifications, real time infrastructure for delivering events as an adjunct mechanism to help improve the experiences within our communication tools, it’s going to be very important. I’m not sure exactly how it’s going to pan out within the context of each of these tools, but I think in the enterprise and for consumers it’s going to be pretty important.
STEVE GILLMOR: Do you think it intersects with Azure in any interesting way?
RAY OZZIE: I think Azure provides a great technology, a great computing and communications fabric upon which those kinds of system can be built, because what you want for scale are systems, back-end systems that can start out relatively small but as traffic gets higher and higher can dynamically expand to serve that kind of traffic, and I think the elasticity of that infrastructure is pretty handy.