Research 2.0 Monthly February, 2007

THE TECHNOLOGY MONTHLY Volume 2, Number 1 February 26, 2007

BOSTON|NEW YORK|PARIS www.research2zero.com

In this issue: The key criteria for judging software in the Web 2.0, SOA world Two emerging platforms: WordPress and Zimbra Update on the data and information management space The Web 20 status and a check on other Web 2.0 adoption metrics

Panning for Software How does one separate the gold from the muck in the software space given the mania surrounding Web 2.0 and service-oriented architecture (SOA)? These terms come with such baggage that it makes it hard to see through to the underlying software. Web 2.0 is so broad it brings in business models, content types, implementation methods, usage patterns and even touches on emotional and social attributes. It’s no wonder the term, so overloaded, defies precise articulation and can no longer be called a definition. Most practitioners rely on the old standby of, “We know it when we see it.” SOA may be a better place to start. However SOA has also been extended in many directions, mostly to suit the needs of corporate computing, so that it too has become impossibly heavy. Multiple standards groups and technology vendors have increased the complexity of SOA to include security, process definition and orchestration, monitoring and event management, application frameworks and so on. The shear mass and level of specification has grown beyond the realm of most potential implementers. An analogy we imagine is that Web 2.0 and SOA are bills voted into law by congress and then burdened with hundreds of additional riders tacked by a broad range of special interests, blunting and obscuring the original intent. We’re looking for the equivalent of a fundamental bill of rights for software without extraneous additions. It’s altogether fitting and proper that we should take the time to do this. It simplifies the discussion and halts the creeping spread of our definitions to encompass things that are well beyond basic principles. The myriad aspects of future software can be distilled down to five fundamental principles:

1. Bind and be bound – Any software function must be able to bind to whatever interfaces and data are made available to it. Similarly whatever functions and data are available must be easily identified and bound to external agents. Note that these bindings are arbitrary rather than a priori. 2. Call and be called – Rigid, hierarchical control structures are anathema to a web of distributed content and users. Control must be taken when required and promptly surrendered when requested. 3. Transparency – The functions available as well as their detailed mechanisms of action and interfaces must be clearly defined and easily understood in a matter that can be trusted. It’s about more than open source. 4. Context awareness – Taking maximum advantage of context creates dramatic improvements in power and usability. The apparent unwillingness of most developers to just exploit what is already possible today has resulted in painfully slow advances. 5. Sub-second response time – Even in a world of rapidly expanding bandwidth we will continue to struggle to meet demands for low latency. Response time is directly tied to major increases in usage. This has implications for architecture and service delivery. So that’s it. No less. No more. In fact taken together the first three might be grouped into what could be called ultimate extensibility. The last two might be thought of as the end-user experience test. What about everything else? There are still plenty of important aspects to Web 2.0 and SOA ranging from new business models to social computing. However they can get in the way when we are aiming at software futures. Many of these are just emergent properties, use cases or implementation styles rather than fundamental characteristics.

Turning back to our five properties we now delve into each one in greater detail.

easily import from other programs but can’t export. The problems with unidirectional design are self-evident.

Binding: Everyone understands the most basic property of binding in that some variable such as is bound to an actual number to be used in a computation. The bindings in the past have been fairly direct in mapping program variables to data. The mind needs to stretch just a little bit to grasp that the new binding involves elements outside of an individual program, and that the binding is not just to data but also to other programs and functions. Data elements, software functions and computing resources are scattered around and come together dynamically as needed to perform a task.

Calling is closely related to binding because binding has to do with functions and data whereas calling manages control and sequencing. If we were real purists we could probably rely just on binding but the control issue is too much of an obstacle not to deal with it separately.

The application programming interface (API) is the mechanism that provides the binding flexibility needed in most situations. It’s important that the definition of the API be granular enough and bidirectional to suffice for any arbitrary use case. The Internet is expanding to include just about every electronic device, including mundane objects such as your stereo, phone, running shoes, picture frame and so on. Often the interfaces may be described but unnaturally limited by the developer. For example an IP phone has a keypad and a speaker that should be usable in any way, not just for entering numbers or playing audio. If we want to bind to the 5 on the keypad and make it display the current status of our traffic variable, it should be simple to do. The point here is data and function bindings can be arbitrary. It’s this property which gives rise to the whole mashup craze whereby users can do things such as link together data from say, Google Maps, Flickr Photos and Craigslist apartments to show a local map of available apartments with neighborhood photos. Thanks to the surfacing of APIs and the use of tags these applications became far more powerful than for a single or small group of use cases. These first mash-ups, while limited, clearly get the point across and have created quite a bit of enthusiasm. Yahoo Pipes is yet another step in this direction.

Calling: It seems that the desire for hierarchy and control can be as difficult to eradicate in software as it is in people. Most modern software allows some level of integration and interoperability but assumes a master/slave relationship. Even casual users have experienced a version of this from software programs that

Calendars are a common example of what goes wrong with calling techniques. By and large every calendar is designed to be the master and have control. They have improved on the ability to import and export basic data but think for some reason that users will enter events directly in that particular UI vs. another, or that they will want to schedule meetings or contact people only in their own database. Because of this, calendar creators have spent much of the last 20 years rewriting the code for “enter a meeting” rather than actually creating some sorely needed new functionality. Zimbra is perhaps a notable exception (see below).

Transparency: This is probably the murkiest and most challenging element of our definition. No doubt some will feel the requirements for real transparency are unobtainable. However even incremental improvements confer substantial benefits and advantages. At a high level users need to quickly understand what functions are provided, what interfaces are available and how the functions are implemented and results obtained. And all that is just to get started. Beyond an initial implementation users become dependent, so any changes can cause ripple effects. People are already experiencing this phenomenon as Google and others evolve their interface support and deliver new releases. These issues are slowing down innovation because most are still choosing to build from the ground up rather than build upon existing functionality. Larry Page of PERL fame has issued a warning that too many developers are focused on creating their own products rather than improving those already out there. Open source is a major industry trend that does help in this area. It’s hard to come up with a more transparent way for software than making source code available to all. But code is not always the easiest source of functional insight for the non-hardcore programmer, especially for complex programs. Furthermore complexity and license terms prevent it from being a universal solution, especially for business. Page ‐ 2

© February 26, 2007 

 

Lastly there remain myriad legal and business issues surrounding terms of use that are likely to cloud this area even further. The insistence on technology such as digital rights management (DRM) when rights surrounding music have always been extremely murky – often split between artists, studios, and songwriters and then again by geography – seems absurd. Thanks to Apple and Google progress is being made in this area, but there is still a long way to go. Stable standards such as HTML, PHP and many others are helping to make functionality and mechanisms of action very easy to understand. Even the simple “view source” ability common in most tools and browsers today has given a major boost to the practice of transparent systems.

Context: Alan Kay highlighted the woeful lack of innovation in this area by noting that most computers are still using the same general-purpose interface that his team designed forty years ago! This despite the fact that it has been demonstrated that productivity skyrockets when users are given an interface more specific to their task. We too are amazed at the lack of sophistication here. After all, our ever-more powerful and connected computers know everything about our history and standard actions. Even within the context of a general UI, why aren’t normal actions anticipated to the point of offering us more one-button options to complete an action? This will be very important in the mobile computing space. In a comic example, one large carrier launched a video feature for users that required eight clicks with a rather high latency to play a video clip. Not surprisingly, users turned their noses up at the option. If you receive an email from a friend (address book, frequent emails) and it includes a video clip why not present a button to simply play it? Why not even pre-fetch it while I am reading the email? Contextual awareness is likely to be a topic for some time to come. Today, even small steps can lead to big results. YouTube videos offer the ever-so-simple play symbol under the user’s mouse button. It’s so easy to click, most do. It might be trivial but we are just at the early stages of seeing things like this at work.

receiving some careful scientific study, but we think everyone would agree that the results are valid whether a user is sharing one machine with thousands or having thousands of machines serving them. Most assume that thanks to Moore’s Law and the bandwidth explosion users will never have response time issues. Think again. It turns out that the addition of computing resources can make the problem even harder to solve. If from France I request a piece of content created in China, we generally have a cached version somewhere closer to home. A simple timestamp tells us if the cache is current. However when you scale up the number of users and sources dramatically so goes the number of cache locations. Figuring out how to serve the right content can take more computing time than just going back to the original source without clever algorithms. The AJAX programming style went mainstream very rapidly because it struck right at the heart of this issue. Thanks to AJAX it was possible to deliver Web-based interaction without requiring the send/receive loop that was once part of mainframe computing and still dominates the Web. Average user response time serves as a gate on the speed of adoption. It’s a real, often overlooked aspect of commercial adoption analysis. The Google founders are relentless at measuring the initial results response for their search. They know that having good results back in less than one second is better than great results back in three seconds. On second thought, Google really is one giant timesharing mainframe so our regular readers are right after all. These five key characteristics define what some call Web 2.0 software. Notice we don’t have to explicitly include implementation details, business models, content types or user roles in our definition.

ÇÇÇ

Sub-second: We expect to hear some groans of, “That’s so IBM mainframe,” from our regular readers. It’s true that timesharing was the catalyst for this area Page ‐ 3 © February 26, 2007 

 

Emerging Platforms Google, Microsoft and Yahoo aren’t the only important players in the ongoing Web 2.0 platform war. Below the surface there is a layer of activity by a number of providers delivering specific functions and expanding rapidly in the market. Both Zimbra and WordPress are emerging as potent contenders for large numbers of users at the expense of the Big Three.

WordPress: It’s becoming commonly recognized that what started out as blogging software is rapidly evolving into a full suite of functions and online services. We migrated to WordPress ourselves a few months ago and were surprised by the broad array of features in its new release. Even more importantly, the community that has grown up around this open source project has allowed capabilities to mushroom far beyond what the core developers could have delivered. Thanks to a modular architecture for the presentation layer (themes) and the internal functions (plug-ins), extensions are easily delivered and installed by endusers. (They have recently added widgets, which fall somewhere in the middle of presentation and functions but we’re not sure how much we believe in the widget approach in general.) WordPress relies on standards such as CSS in the presentation layer, which means users can easily grasp and customize the look and feel to suit their needs. Elsewhere, WordPress adheres closely to appropriate standards and has foundations on common denominator technologies such as HTML, PHP and MySQL. It’s hard to do justice to the velocity of the WordPress advances except to say that they make Google seem like a very slow software company by comparison. Recently the company has decided to embark on a more regular release schedule of every three to four months. The company invites users to submit and vote on ideas for features to be included in subsequent releases. As of this writing there were 422 ideas submitted and more than 24,000 votes on which should get the highest priority. This effectively expands the development direction beyond code submitters to general user community and enhances the transparency of the platform. Spam is a scourge on everything these days, blogs included. The company has developed a tool called Akismet, which is a very good spam filter. Thus far we would say it is more effective than the filter used by Google for GMail. Akismet is critical as well. More

than 90% of blog comments are spam and would quickly choke off and eliminate the commenting feature if not stopped effectively. The 2.0 series release of WordPress garnered 1.8M downloads compared to the 207,981 downloads of the 1.5 release on May 9, 2005. It’s clear from these and other statistics that the growth of WordPress has taken off and is likely to maintain a fast pace. Automattic, the company behind WordPress, boasts just 11 staff as of this writing and received a minor amount of funding from a few venture capital types. WordPress, both the software and the company hosted solution, is available for free. The company does make some money via its antispam technology, services and consulting. Support is offered at $2,500 to $5,000 per company contact similar to other open source approaches from companies such as MySQL. The other major blogging platform providers are Six Apart (TypePad) and Google (Blogger). Six Apart has been aggressive but growing a bit more in a MySpace sort of direction, which is less interesting to us. Since Blogger was acquired by Google its pace of innovation has been slow. Google seems far more focused on Docs & Spreadsheets as a platform play.

Zimbra: Turning more directly to the corporate

space, there is finally some meaningful innovation around enterprise messaging and collaboration. Microsoft and IBM Lotus have enjoyed a duopoly here with little or no competition for some time. Zimbra is a next-generation platform that exploits new technologies in a product built from the ground up to understand context and integrate diverse functions in an intuitive fashion. With a recent user count of 6M the company has certainly reached a critical mass. At first the application looks very much like just another open source implementation of a standard product (Microsoft Outlook), but that’s where the similarity ends. What’s different about Zimbra? To date Zimbra is the best example of an application that makes use of context that we have seen to date. Users can get information and take actions from anywhere rather than navigate to and from various task screens and applications. As with many ideas the value is in the quality of the implementation. For example, email content is parsed and recognized so that a hover or click produces logical in-place results. It’s worth watching the Page ‐ 4

© February 26, 2007 

 

flash demo available from the company website to appreciate. Of course we have seen good collaboration suites come and go over the years with little adoption. Zimbra has been smart enough to recognize that they are a newcomer and support full integration with Microsoft Outlook, Blackberry and other popular email clients. The administrative back-end has received the same level of attention in design and ease-of-use to help overcome enterprise inertia. The software is also designed for service providers who offer Zimbra collaboration as a hosted (SaaS) solution. Enterprises can easily extend and integrate [Zimbra directly into their workflow with build in tools for creating “Zimlets.” We may not be fans of the name (sounds too much like Gimlet or the ill-fated Zaplet to us) but the implementation and pre-defined Zimlets are powerful and intuitive examples of a fully-extensible collaboration suite. A few examples:

For those not wanting a pure open source solution there is a supported version of the product priced at $2030 per user per year. There are also many providers of the software as a hosted solution where the prices range broadly but get as low as $50-75/user per year; fairly similar pricing to Google’s version of office. Recently Zimbra crossed the six million user mark, clearly putting it on the map of enterprise readiness. The company has substantial venture capital backing from large investors including Accel, Benchmark and Redpoint. WordPress and Zimbra are strong examples of nextgeneration software and infrastructure for modern webbased computing. They may not be making headlines yet, but their current size and trajectory implies that they will.

ÇÇÇ

1. Integration with VoIP solutions from Cisco, Asterisk and Skype. Especially for companies considering Asterisk, Zimbra greatly enhances the usability for the end-user. 2. Links with Salesforce.com to allow automatic updating of CRM information directly from the collaboration windows. This helps eliminate the mental overload of having to enter data twice. 3. Hooks into travel applications for booking, tracking and mapping plans. We’ve been amazed that it has taken so long to begin integrating travel functions and information with schedules and contacts given that it is a fairly high percentage of business related activities. Finally a user can drag an event invitation and drop it on an icon to initiate a travel search and populate the calendar with those travel arrangements automatically! The real power for enterprises will come from creating their own extensions to address their basic business processes and integrate them right into the collaboration flow for their end-users. White collar productivity is notoriously hard to measure but a full implementation of Zimbra with some customization would offer a meaningful 10%+ increase in efficiency and effectiveness.

Page ‐ 5 © February 26, 2007 

 

Web 20 Update and Other Barometers

out, the group could easily add another 10-20% from these levels for the year.

During the past year the Web 20 has traveled a winding road basically to get back to where it started. Company Akamai Equinix WebEx Intuit BEA Google Apple F5 Salesforce Adobe Amazon Yahoo NAVTEQ Red Hat QUALCOMM eBay Broadcom Websense Rackable SanDisk

Figure 1 – Web 20 TTM Performance

Also buried in the flattish 12-month performance was a large divergence within the group shown in Table 1.

Group Total

YoY Price Δ 157% 79% 53% 20% 19% 16% 14% 11% 7% -2% -16% -18% -21% -21% -21% -25% -30% -34% -36% -40%

YoY Rev Δ 51% 30% 23% 19% 11% 73% 38% 40% 60% 31% 26% 23% 17% 32% 33% 31% 37% 20% 67% 41%

5.5%

35%

Table 1 – Web 20 YoY Changes

2008 P/E 35x 13x 19x 19x 17x 27x 22x 29x 63x 23x 38x 39x 23x 31x 18x 22x 31x 18x 13x 14x

2008 PEG 0.92 0.46 0.92 1.46 1.00 0.55 0.90 0.89 1.33 1.17 1.88 1.93 1.17 1.00 0.85 0.90 1.46 0.94 0.31 0.50

4.6x

26x

1.02

Table 2 – Web 20 Valuation Ratios

A full table of metrics for the Web 20 and a dynamic version of our Web 20 chart available as a Google Widget can be found at the Web site.

Skype This month Skype downloads crossed the half-billion mark. Our usage data for Skype shows continuing adoption. Given the variation around both time of day and holidays it is clearly being driven by business use. Skype Peak/Avg. Concurrent Users (Aug '06 - Jan '0 9.5

Peak CCUs

8.5

Peak CCUs Trend Avg. CCUs Avg. CCUs Trend

7.5

6.5

5.5

4.5

3.5

6

6

07 nJa

-0 ec D

-0 ov N

6

6

6 -0 ct O

0 pSe

-0 ug A

Average estimates for 2007 results suggest revenue growth of about 26% for the group. We are starting the year about 15% higher in terms of price. When we roll our numbers forward to begin to use EV/Revenue multiples on 2007 estimates and P/E on 2008 the valuations improve substantially. If estimates prove

Group Total

Concurrent Users/Millions

Company Akamai Equinix WebEx Intuit BEA Google Apple F5 Salesforce Adobe Amazon Yahoo NAVTEQ Red Hat QUALCOMM eBay Broadcom Websense Rackable SanDisk

2007 EV/S 15.3x 7.2x 3.3x 3.7x 2.6x 9.4x 2.4x 4.7x 6.6x 6.9x 1.2x 6.9x 4.3x 8.1x 6.4x 5.8x 3.8x 3.0x 0.7x 1.8x

Figure 2 – Skype Usage

Page ‐ 6 © February 26, 2007 

 

There’s clearly a minor divergence between average and peak Skype users. This suggests an increasing proportion of users who login for a specific reason rather than those who stay online all day inviting calls.

Job Listings We’ve started to take some snapshots of job listing data to begin to build a useful time series. Just for fun here are some of the numbers for the current month. (All are from the Simply Hired job search engine, which had 4,747,436 listings at the time.) Jobs by OS (153,892)

Linux 22%

Solaris 10%

Windows 68%

and applications for using information; and 2) Volumes of data and information continue to grow at an accelerating rate. Along with the increase in volumes and types of data gathered, the cycle times available for processing and decision making get shorter; substantially compounding the problem. Every few years this reaches a sort of threshold, prompting the active consideration and adoption of new applications and approaches. Investors are likely to revisit this area thanks to the coming spin-off of Teradata from NCR, creating a new standalone company with more than $1.5B in revenues. After the spin-off NCR will be more of a pure play on retail POS technology with over $4.5B in revenue. NCR stock has been on a tear to a projected fair value of $47/share based on an estimate of the value of Teradata and NCR as separate businesses. The table below suggests that post spin-off NCR may still have some upside. If it can execute its way to an average EV/R multiple the stock could be worth 2x current prices.

Figure 3 – Operating Systems

Windows still occupies the lion’s share of hiring activity but we will see how this develops during 2007. Jobs by Language (125,125)

AJAX 5% C# 20%

Diebold VeriFone Micros TOTAL EMC NCR

EV

Rev

OI

EV/R

R/Emp

$3.5B $3.5B $2.3B $9.2B $31B $8.5B

$2.9B $581M $679M $4.2B $11B $6.1B

$153M $108M $91M $352M $1.2B $473M

1.2x 6.0x 3.4x 2.2x 2.8x 1.4x

$199K $254K $179K $203K $415K $216K

Perl 17% PHP 6% Ruby 1%

Java 51%

Figure 4 – Languages

Each month we will be expanding this section with more data as we collect more and add new external sources.

ÇÇÇ Update: Information Management Corporate interest in information and data management technologies is very high. This is likely to remain true for some time for two very basic reasons: 1) There seem to always be new high-return opportunities

Table 3 – POS and Data Company Figures

Teradata may be better off outside of NCR, but it’s hard to imagine that the company will be able to stay independent. Outside of financial ratios, the firms that own the data have a special status. Teradata has a market position that is just too tempting strategically for firms from Microsoft to HP to EMC. Exploring the vast data resources has long been the purview of large established players such as Oracle, Business Objects, Cognos and others. A frequent question these days is whether these existing solutions are passé and whether new offerings from small companies such as Greenplum, Endeca, DATAllegro, RiverGlass, SeeWhy, Dabble DB, Kaskad and StreamBase reflect the future direction of data mining and business intelligence. It’s certainly true that new approaches to processing real-time data coming from the Web or other sources are garnering more market share. This could be eviPage ‐ 7

© February 26, 2007 

 

dence that information management is entering a new phase where customers are seeing distinct application areas differently. It could be the case that we will see sustained differences in how customers approach strategic vs. operational BI. This could mean that small areas today could grow to challenge the data and information management Death Stars of Oracle, EMC and IBM. Some of these new approaches will catch on, and the large players will acquire or replicate their way to offering similar capabilities. Between now and then, however, the market is offering opportunity to grow into critical mass ($25-50M revenues) and even reach potential public market revenue levels of $100M+. It’s too soon to tell if it will be possible for any company to propel itself into the multi-hundred-million-dollar level of sustainability. For every success (Business Objects, Informatica, MicroStrategy) there is a failure (Pervasive, Versant, Persistence) or one that was absorbed (Informix, Brio, Crystal) by an existing player. The overall landscape of public players in the data space paints an alluring picture as even companies that institutions might consider also-rans (Sybase, Progress, Hyperion) trade at substantial market values. Company Oracle EMC Sybase Progress Bus. Objects Cognos MicroStrategy Informatica Actuate SPSS Hyperion Fair Isaac Open Text Applix

2007E Revenue 17.6B 12.7B 1.0B 471M 1.4B 972M 313M 374M 137M 290M 890M 864M 584M 67M

Enterprise Value 87.4B 30.9B 2.2B 931M 3.1B 3.2B 1.5B 910M 255M 492M 2.0B 2.5B 1.4B 148M

2007 EV/Rev 5.0x 2.4x 2.2x 2.0x 2.2x 3.4x 4.1x 2.4x 1.9x 1.7x 2.2x 2.9x 2.4x 2.2x

Group Total

37.7B

137B

3.6x

Table 4 – Data and Information Company Values

pany gets there is not only half the fun but also perhaps two-thirds the final valuation! We expect many of the existing emerging players in this space to do well and others to emerge over the course of 2007 thanks to increasing awareness of the potential.

THE TECHNOLOGY MONTHLY Kris Tuttle, Editor Pete Bishop, Analyst Charlotte Borde, Copy Editor The Technology Monthly is published by Research 2.0 12 times a year by Blue Caterpillar LLC, 1313 Washington Street, Boston MA 02118. USA Telephone: 617.381.4762. Individual subscriptions by the month or year are $32 and $256 respectively. Some free content is available via email and our blog. Please visit us online at www.research2zero.com. It is a violation of federal copyright law to reproduce or distribute any part of this publication to anyone by any means without permission (often readily provided.) Sponsorships come with internal redistribution rights and in some cases the right to distribute externally to customers and prospects is available. Copyright 2007, Blue Caterpillar LLC. All rights reserved.

Coming in future issues… Gödel would be short Proof of Moore’s Law for Energy More emerging platforms What happened to .NET? Virtualization tornado? Update: Enterprise IT More Web 2.0 adoption metrics Individual company reports

In conclusion the data and information space offers real opportunity for new players. Other areas such as virtualization are hot but don’t offer the freedom of entry and differentiation that is needed to create major new value. There is clearly still some art to be had in that getting to $300M in revenues can result in a market valuation ranging from $500M to $1.5B. How a comPage ‐ 8 © February 26, 2007