The highlight of JavaOne for me has become supper with Gianugo Rabellino, the founder and CEO of SourceSense. For now, each year…

"'The time has come,' the Walrus said, 'To talk of many things: Of shoes—and ships—and sealing-wax— Of cabbages—and kings— And why the sea is boiling hot— And whether pigs have wings.'"
-- from the poem "The Walrus and the Carpenter" within Through the Looking Glass by Lewis Carroll

… For our conversations are wide ranging and thoroughly engaging as we indulge our enjoyment of fine food and open source.

This year, Gianugo has been in the USA several times, for the Open Source Think Tank and the Open Source Business Conference, and on business. An employee of SourceSense even had a gig in the town in which I was raised. Unfortunately discovering it to be the arm pit of the United States. Ah well, it's why I live in the SF Bay Area now.

We talked of tempering chocolate, unusual eating habits of various cultures, the worsening economy, European vs USA political views, and Gianugo even had me read, while sober, the Pronunciation Poem. [I flubbed a few words, and disagree with some of the rhymes given that my dialect is as Philly as a Chesse Steak.]

But we also talked about Open Source; all the open source related conferences and blogging and work and newsworthy activities of the past few months. Here are some of my thoughts.

Open Source is a Philosophy

This one has been bubbling around in my head for years. Open Source, in and of itself is NOT

• a licensing template
• a business model
• a development methodology

It is a philosophy that can provide the framework for those three things with which it is most often identified. The variations among what the open source philosophy means to each of its followers can best be seen from the proliferation of licenses, business models and methodologies all claiming to be open source.

To me, the open source philosophy is very simple, and it can be applied to solutions for some very complex problems and concerns: the source is available to anyone who obtains the end-product, whether one has obtained the end-product through a no-cost download or through a purchase agreement of whatever type. The "source" may be the "source code" for software, but, to me, it should be whatever specifications and design documentation are required to recreate the end-product in either the original or a modified form.

There are a variety of reasons that a developer, an engineer, an inventor, a creator, might what to subscribe to an open source philosophy. And each individual or business must decide if those reasons make creative and economic sense for them. Once one has decided to subscribe to open source, or any philosophy, then all your other decisions will only lead to success if they are logically consistent within that philosophy. The licensing language, business model, economic forecasting, internal processes and external relationships should form a coherent whole within the underlying philosophical framework.

Business Models Will be Different for Different Markets

Organizations often look for the silver bullet or the golden mantra or the platinum ring that will solve their problems, lead to dominance in the marketplace, or allow them to rule them all. So we hear a lot of talk about the "best" open source business model. This search ignores that fact that there are open source solutions for many markets. Without going into specific verticals, let's just consider four general, horizontal markets that are addressed by open source software.

1. Information Technology Infrastructure: including operating systems such as the various flavours of linux and BSD unix, application servers, and other middleware such as the Mule ESB, WSO2 SOA solutions & KETTLE for ETL, web servers, email MTA, the Funambol Mobile Server platform, and many more.
2. End User Applications on the Desktop/Laptop/Mobile-Device: with OpenOffice.org and its MacOSX offshoot NeoOffice, Projivity OpenProj, and Mozilla Firefox and Thunderbird likely being the most well known, but also including many thousands of open source projects for artistic creation and enjoyment, as well as office and personal productivity, and gaming, and all things that one personally does on a computer
3. Software Development Tools: from the Eclipse IDE to hundreds and thousands of specific language libraries and everything else a developer might need; this is even more "by geeks for geeks" than the infrastructure area
4. Enterprise Applications: these are end-user facing applications, often fulfilling mission critical needs, such as ERP, CRM and BI, and is the least mature market for open source providers and the hardest area to gain acceptance

No one business model, no matter how generically expressed, could satisfy these four disparate markets. Monetizing these areas will come from combining innovative and traditional packaging of support, customizations, system integration, training, licensing, and subscriptions.

Automated maintenance, repair and update networks might work very well for monetizing IT infrastructure, but might be insufficient for the other markets. Ad based monetization, directly or through partnerships with ad networks like Google's, might work for some end-user applications. Putting out the begging bowl, asking for PayPal contributions might also work. But for many open source projects, there simply isn't any path to monetization.

One area that I think has been insufficiently explored, and might well be the only path to success for the Enterprise Application open source vendors is Software as a Service. The SaaS approach, whether through partner channels or directly, is the most sensible means of monetizing a wide variety of open source applications. Embrace the open source philosophy, leverage the strengths of flexibility and community in the licensing, business models and processes, and monetize through SaaS delivery into vertical and niche markets.

And for the rest, just acknowledging that your company is a software company embracing an open source philosophy, and building appropriate support and licensing structures will be the best path.

Communities are Strongest when Open

Developing a community around a product is not unique to the open source world. Microsoft, Oracle, SAP, SAS, SPSS, IBM, etc, etc, etc have all developed and supported great communities made up of ISVs, VARs and Users. The difference however, is that open source projects are very dependent on their community. Conversely, communities are strongest when there are no artificial limits to communication.

I think that most companies are embracing at least this one ramification of the open source philosophy. Though it seems to me that when the source isn't open, the community can only converse and support each other in surface issues.

One area though, where open source vendors still seem to be uncertain, is the definition of community, and what groups make up their community, and if they must develop different communities for different categories of members: developers, users, non-software contributors, paying customers, or partners. I think this is self-defeating. There is only one community for each project: those who want to be involved and will benefit from that involvement, ultimately to the betterment of the project, whether directly or indirectly, whether monetary or not, whether contributing code or documentation or use cases or testing or playing around with the project or complaining about it or just soaking up the ambience. Can any open source project community manager give a valid reason why any of these should be excluded from the community or ignored?

Open Source Sales are Very Foreign to Most IT Buyers

Whether we're talking about senior IT management, the CIO, or the purchasing department, the idea of freely downloading the fully operational, unlimited version of a product, and using it for research, prototyping or production, without handholding, cosseting or sales incentives is complete anathema to those making purchasing decisions.

Did you ever see a "vendor bake-off" among products from Oracle, IBM and SAP without vendor involvement, without vendor sales engineers fine-tuning and providing weeks of free labour? Is there any open source company that provides a corporate jet and box seats at the super bowl? And is there anyone who has ever been part of selling (internally or as a vendor) a large, big-budget enterprise project to a CIO or purchasing committee, who doesn't believe that such freebies and perks are essential elements to the approval process?

Many, too many, open source vendor CEOs believe that bloatware (more features than most users will use), vendor lock-in and lack of interoperability, and high prices of initial licenses are the problems CIOs worry and that open source software solves. They think success will come from "80% of the capabilities for 20% of the cost" of their closed source competitors. I don't believe any of it. CIOs worry about being beat up by the business side and greasing the squeaky wheel. The two highest costs in a data center are people and energy, not licenses or maintenance agreements. Implementation costs: hardware, software licenses, personnel and training, are spread out over three to five years with appropriate tax implications. For really large projects such as a full ERP or data warehouse, there is at least a ten year life expectancy. Why do you think that there are so many mainframes and COBOL applications still around? The implementation costs were written-off long ago, and the on-going support costs are minimal because they still JUST WORK. Those with an eye to economics are more concerned about the total cost of ownership over the complete 10-year lifecycle of a capital project, than with the 10% or less of those costs that go towards initial software licenses.

So, if the only way for open source vendors to become big players, or as one CIO put it, to "move out of the junior varsity", is to look like the big players, to provide pre-sales engineering and collateral beyond a T-shirt, then what will this do the business model and the belief in "20% of the cost"? It will destroy it. This is why there are more openings for sales & marketing than for engineers at open source companies today.

Open Source Vendors will grow and evolve and come to operate more and more like the big players they're destined to become. As long as they continue to adhere to their form of the open source philosophy, are internally consistent and true to the logic of their open source framework, they'll continue to benefit from the true value of open source: flexibility and community knowledge to a greater degree than can be achieved by any closed-source vendor or any isolated, enterprise data center with home-grown solutions. Speaking of which…

Open Source is the Middling Ground between Build vs Buy

This is another point I've been hammering for years. Traditionally, IT shops were either build or buy. The build shops were very development oriented and created custom solutions to implement business processes and support business users. The buy shops bought COTS software, and either convinced the business that the software they bought implemented industry best practices, or spent millions of dollars in customization.

Build shops suffer from isolation and the total burden of maintaining and updating the software they built. When the only developer left who remembers and understands the code, retires, faith often becomes the best practice for ongoing support.

Buy shops suffer from ongoing dependancy on the vendor(s) and ongoing compromises for the users.

Build shops should be (but often aren't) more flexible to respond to changing business needs, and can be more valuable to the business.

Buy shops can be more reliable and cost-effective - really. But not always.

One result of Y2K [remember that?] was that most IT departments became much more of a blend of build and buy, with buy decisions winning out. Coupling this with the facilitation of distributed workgroups made possible by the Internet expanding world wide, and offsourcing [outsourcing operations to offshore companies] decimated many corporate data centers by solving the personnel and energy cost problems. Guess what? Implementation and licensing and maintenance costs remain the same. Some CIOs use offsourcing as the reason that they don't use open source; the outsourced vendor isn't familiar with it, and has no incentive to use it.

As companies bring IT back inside, and as its importance to the business is once again realized, open source offers a third path to the traditional build or buy. However, offsourcing will still be a substantial part of the solution, and should lead OSVs to recognize the importance of offsourcing vendors and embrace them as partners and very important channels.

The IT shop that responds to the business using open source can be both flexible and well-supported. The open source vendors that make such IT shops succeed through flexibility and reliability will be the most successful ones. This can be achieved through the OSVs growing their support, professional services and training organizations, or by partnering with all sizes of PS, VAR and outsourcing firms, or both.

Conclusions

The real conclusion is that bringing the open source philosophy to fruition in business is still an evolving process. The advantages gained by the openness of the source and the strength of the community is being recognized by both the IT shops and the OSVs. While there are cost savings to be had, OSVs need to stop relying on initial licensing cost reductions as their main selling point, and begin to market the advantages to the IT shop of using open source: responsiveness to changing business needs and increasing reliability over time, all while providing the best TCO and ROI.

There's a lot more to be said on all of these topics and opinions, and maybe I'll even get the time do so.

Startup Camp is a mix of traditional conference panels and an open space "unconference".

The juxtaposition of the two is a bit jarring in some ways. 'Tis somewhat like going from a warm fuzzy blanket [panels] to an exhilarating plunge into the chilly North Pacific [unconference sessions]. Actually, panels have been losing their appeal over the past few years. A good panel is a lively debate. The panelists engage each other. A good panel is both entertaining and informative. Too many panels today bring to mind an event from my college days. This is an unfair analogy but I'll tell the story anyway.

The only diner within walking distance of the college was the S&W [or, as it was traditionally known, the "Slut & Whore" - hey, don't yell at me; it describes the area well. Think of the Tenderloin in SF on a bad night.] Anyway, after an all-night study session, or maybe a drunken game of Diplomacy, we made a visit to the S&W in the wee hours for some coffee and rice pudding. Among the other patrons were two old transients [a.k.a. hobos, stumblebums]. They were at opposite ends of the diner, but seemed to be having a loud argument on who was the greatest baseball player of all time. After a while, we realized that they weren't arguing. They were merely proclaiming in loud voices on the same subject. OK, the panels aren't that bad, but getting close.

Another problem with panels today is that most panelists blog. If you are at all interested in the subject, it's likely that you read their blogs and already know their opinions.

Ah well, onto the good part.

The unconference sessions are intimate and, as mentioned before, exhilarating.

The only problem is that there are so many great topics being discussed at the same time, that it's like being in a candy store: you don't what to grab onto, there are just too many choices.

But you can read all about what happened on the Startup Camp wiki, if the attendees upload their notes. Just follow the links from the Startup Camp Unconference schedule at the bottom of the page.

Tomorow will be running back and forth between Startup Camp and CommunityOne. Can't wait.

Today, I'm at Startup Camp, in conjunction with CommunityOne, which starts tomorrow, in conjunction with JavaOne. The schedule is online, but this is a camp, an unconference, and the rules are different.

The introduction is over, and the keynote with Jonathan Schwartz and Om Malik is going on now.

A question from the audience, essentially that there is no simple, online solution for a Java developer to just go online and develop, led to an interesting side discussion where I'm seated about what it means to be a developer. The old folk at the table, have a much broader definition of a developer, and the skills they should have, than the young'uns, who are more focused on just one language, and just code monkey banging away.

The first panel is starting up, moderated by S. Neil Vineberg, President, Vineberg Communications. The topic is branding, and how the brand comes out of the founders and the culture that they create.

One interesting point for me, is that while the panel is discussing the importance of branding, and that the brand can flow from the founders' personality and the culture, I have found that if the founders are too focused on creating a brand, and selecting their brand category, they'll fail to infuse their brand with their personality, that is, the brand can seem sterile and contrived. Branding is very important, but, to me, it must come naturally from the company if it is to truly reflect the company. Of course, sometimes you have a founder that you need to keep locked in a closet, feed caffeinated drinks and pizza, and never, ever let them talk to the customers.

The use of social media is a given today, especially for startups. The use of Ning [why doesn't PeopleAggregator ever get a mention?] and the growth of social network platforms for startups [others] to grow their own social networks is a great indicator of this.

For all the talk of social media, the message of this panel seems to be that traditional methods, through PR, is still the best way to reach out and get your message across; especially outside of the technology centers like the SF Bay Area. However, it seems to me that when the panel starts talking about what's really effective, they use terms like "community" and "authentic conversations". It also seems that creating markets and driving markets is the better than pushing a product, in terms of success, and in terms of getting noticed.

The morning is done. I'm looking forward to networking and learning throughout today and tomorrow, to CommunityOne and seeing Michael Coté at Redmonk's Unconference, and a great conversation with Gianugo Rabellino at JavaOne and for supper on Thursday.

InfoWorld has made downloads available for selected presentations from this year's Open Source Business Conference. The links will take you to PDF files. But, Matt, where are your video mashups?

The community has been brought together for the final campground session. Rather than sing kumbaya, we've been teased with t-shirts, toys and books from O'Reilly and being treated to a demo of a hot deployment by Travis Carlson. 'Tis all drag-and-drop goodness, with dragging jar files (bundles) as connectors and apps from a MacOSX Finder into Mule and watching the results in Terminal.app. Ohhh! Ahhh! Safaris shows Mule saying "Hello Travis, welcome to MuleCon" OK, I'll admit that this is pretty neat and should satisfy the most addicted user of Tibco wizards. This won't appear in Mule Enterprise Edition until after the 2008Q3 release of Mule EE 2.0.

One little aside… I was talking to an ex-TIBCO employee who said that as old as TIBCO is (grew out of a Teknecron business that was founded in 1985 and became an independent business in 1997), the user even only draws about 400 people. This is only the second MuleCon [there was a sort-of pre-MuleCon in 2006 at a bar in London, but let's not count it), which last year drew around 100 folk and this year brought in around 240. That's phenomenal growth and shows the excitement that good software can bring to its users.

Questions for the Campground from Day 1 and some new ones.

1. AMQP is an emerging messaging protocol and is platform independent. To answer the question about dot-Net integration with Mule, this is part of the answer.
2. BPEL & BPM? Workflow tools, such as JBPM, that don't depend on simple web services, work better with Mule. Travis has done some work with this including the JBPM plugin that is in the current distribution of Mule, and there is more on the Forge. Mule also works well with Oracle and Iona's solutions. Mule Services and Mule Events are complementary.
3. Mule is moving to Spring2, and has MuleSource thought about the exciting things that can be done with AOP in Spring2? Not so much yet, but it is among the next thing to look at. Mule Galaxy might be one of the first areas where MuleSource would leverage AOP to create runtime governance.
4. Can Mule Galaxy can be used to manage other artifacts? The short answer is yes.
5. How are folk clustering Mule? The typical way is to leverage an existing clustered app server. They are seeing more where the ESB has a state itself. They're looking at things like TerraCotta for such use cases.
6. Discussions of rules engines and routing, but much work to be done, hopefully through the MuleForge.

Dave gave some closing remarks and that's all 'til next year.

This first user track session is a panel, once again moderated by Michael Coté of Redmonk with John Davies, Technical Director and Head of Research at Iona, Eugene Ciurana of Leapfrog, John Rowell, CTO of OpSource, and John Gardner, Principal Consultant at MomentumSI. My take away from this is that Mule, perhaps open source in general, allows a company to better balance their risk as they test out the potential rewards while implementing tactically against strategic goals. A great metric from Eugene is that the most wait time he has ever had with Mule for a support issue is 90 minutes from opening a ticket to getting some sort of answer, and this is what keeps selling him on Mule vs proprietary solutions. Cost is a factor in ROI, and Mule wins out on not just initial costs but also costs less in need for training, on-going support, consulting, and maintenance; but beyond this, the flexibility of Mule and ability to integrate it with the infrastructure, and the openness not just of the code but in the company and personnel also adds to the return, as does the "cool" factor.

Having now seen Coté moderate two panels now, I have to say that he's one of the better moderators I've seen.

The second session is "A Quantum Leap in Ease of Use: Introducing Mule IDE 2.0". First let me say that having a nice, GUI IDE will help with the ROI perception, in that it is often the perception that development can be quicker and more efficient if done in a nice, pretty GUI. I happen to disagree, but I'm old and like CLI and typing more than clicking. Let's see what these guy's say. Yep, yep, yep… they're showing a pretty drag-and-drop GUI configuration editor. I took the one-day training at last year's MuleCon, and found that configuration through editing XML files to be very easy, and to help with understanding what Mule was doing. There was a contingent of attendees who complained that they didn't have their Tibco Wizard. This should help placate that contingent. Ah, the Mule IDE 2 is pre-alpha, though you can get it on the Forge. The release is expected to be in 2008Q3. The demonstration of the Mule IDE 2.0 plugin for Eclipse looks good; in addition to the GUI, one can still see the underlying XML.

John Davies, Technical Director and Head of Research of Iona spoke on Financial Messaging in Mule. The concept of losing a message in an investment is literally unthinkable. One interesting tidbit is that MS Excel is the only MS products one finds in an investment bank. John went over the front/middle/back office requirements, and the place where Mule is found is in the middle where XML is heavily used, usually over MQ & JMS. Using Artix, Mule, and GigaSpaces on an Azul box, provides essentially linear scaling and throughput is so high, performance is so great, that the need for transactions and supporting transactional relational databases is eliminated, thus increased performance allows increasing performance even more.

The final session of MuleCon2008's second morning is a panel based upon "Are you Smarter that a Fifth Grader", moderated my Michael Coté of Redmonk. The panel consisted of Dave Rosenberg, Larry Augustin, Jason Maynard and Matt Asay.

The first question was what gets these folk excited in the shower in the morning: SaaS, Web2.0 enterprise/consumer Mashups, or other newly hyped trend. SaaS and mashups seem to be joined in the answers as a way for business users to get around IT by buying what they need on a credit card and expensing back this "rogue IT" service. The question evolved into "Is open source a given?". The answer was "yes" and the discussion segued back into how all these things, SaaS, mashups, open source, allow business users to get what they need quickly. Coté then asked if this meant that there was a "work life retooling" going on that was allowing traditional IT departments to work with Rogue IT implementations. While the consensus seems to be that IT departments were indeed going through the work life retooling to allow Rogue IT implementations, such as with WiFi hot spots or open source, I have to disagree. Every IT group with which we've worked has many regulatory, compliance or responsibility issues that preclude them from allowing Rogue IT. However, if a technology shows its value, even if that value can not be quantified in simple cost-benefits analyses, that technology will be brought into the organization under the IT umbrella, one way or another.

The floor was opened to questions from the attendees.

I asked if the panel would discuss MDM being useless without SOA and vice-versa. They pretty much declined, but Jason did discuss some of his own experiences. The bottom line is that the concepts are difficult, and are only starting to be embraced by developers and companies.

Another question asked why there seems to be so little uptake of open source ERP. I would like to ask the same about OSBI, but didnt' need to do so, as Matt brought it up. At this point, open source seems to fill niches within a given enterprise, and one can build a niche OSBI solution more easily than a niche ERP solution. Larry also pointed out that many large organizations in the US have ERP implemented, and they're not going to rip it out to go open source; however, if you look at other markets, where ERP implementations didn't happen [as part of the Y2K fixes], you can find uptake of open source ERP.

Another attendee pointed out the only the oldest dinosaur can question the quality of open source, yet one still must go underground to implement ERP in most organizations. The panel discussed the difficulties of getting through procurement at a large company, and how tactical implementations [back to rogue IT] rather than strategic are often more plausible. It seems to be a matter of market education, and that takes time. Matt also talked about the post-procurement process, and how that is made easier by open source in that the risk is much less.

BPM, SOA, real-time-enterprise… This nirvana, or any other TLA, doesn't really exist in any enterprise, and many of these concepts are hard sells, whether implemented with open source solutions or not.

The final session for today goes into System Management of Mule and the the additional products that mitigate risk in SOA deployments. This was definitely the most heavily attended of the user sessions.

MuleHQ is a dashboard that provides device-level management of app servers, databases and other assets. Policy driven alerts that allow management against the SLA is also provided. The server side performs auto-discovery of services and assets of which Mule is aware. Mule components, connectors, routers and endpoints are monitored by MuleHQ. MuleHQ is still Hyperic.

The presentation on Mule Saturn started with the challenges of integration, as well as supportability and maintainability of the integration solution. Mule Saturn provides a reduction in discovery of problems and recovery from them to allow meeting SLAs by "instrumenting Mule" to allow increased monitoring and proactive response through Mule Saturn and MuleHQ. Mule Saturn provides a view into the integration process and data flow for business users, whereas MuleHQ provides monitoring for the technical users. There is role based access to allow editing of the XML only by designated users.

The first use case provided was for a Supply Chain Integration. The examples considers a batch of POs containing two orders, one that succeeds and one that fails. Drill down through hyperlinks is possible. Essentially, the use case walked us through Mule Saturn.

The second use case went through the workflow of approving an employee's expense report with a similar walk-through of Mule Saturn showing a failure in the process where the manager never approved the expense resulting from an email failure to communicate.

Ron Park, VP of Engineering at MuleSource, gave a comparison of the Community and Enterprise editions of the Mule products starting with Mule 1.5 then Mule 1.6.

Ron also introduced Jackie, who is in charge of technical publications, and Scott Yuen, who heads support and services. These two areas are of noted improvement over last year [when neither really existed] and what is being made available today to Enterprise customers. In addition to formal documentation and support, there is a knowledgebase available to subscription customers, which contains over a thousand articles.

MuleHQ 3.1 new features include:

• Improved mule specific statistics
• Dashboards including patch tracking through the metadata
• Mass deployment management
• Local data storage of metrics

It should be noted that MuleHQ is intended for stand-alone Mule implementations, and not when Mule is embedded in other SOA tools.

Mule Saturn is included in the Enterprise edition to provide data visualization for a light-weight business activity monitoring service.

Mule Enterprise 1.6 should be out by May, 2008. Again it will include MuleHQ and Mule Saturn, as well as some premium connectors for JDBC, Apache CXF and native WebSphere MQ support.

Mule Enterprise 2.0 will have even more documentation [see Jackie] and the major difference will be the Migration Tools for Mule Core. There will be more premium connectors, such as for FTP, and improved reconnection strategies. Improvements and architectural enhancements to Mule Saturn and MuleHQ are also part of the roadmap. MuleHQ will also have changes to the plug-in architecture, moving forward towards hot deployments.

The community version of Mule 2 is available now, but there will be maintenance releases to Mule Community edition 1.4.

There will be features that will be in Mule Enterprise that won't be in Mule Community. Hot deployments and clustering/fail-over are examples of such features. How these features will be licensed is still being investigated with the lawyers. Mule IDE is open source and available, as is Mule Galaxy for governance, but the two of the working together are an Enterprise feature add.

As MuleSource evolves, I think the differences between the Community and Enterprise editions, the licensing, and the ability of the community to add enterprise-like features to the Community edition will be very dicey. We'll have to wait and see.

OpSource provides web operations for SaaS companies. They deliver about 250 on-demand applications involving billions of transactions per day. The backend is Mule. OpSource using the best solutions to their challenges, open source or proprietary, which says a good deal about Mule.

Their opportunity is DATA. By creating a canonical data model, they have a single set of SQL and XML schmas for the data model resulting in the ability to create ODS and DW. The other opportunity is SERVICES for their customers and their customers' customers.

The solutions was to implement and ESB for Integration/SOA, which resulted in a fully multi-tenant version of the Mule ESB. OpSource Connect is reliable and scalable, and fully PCI compliant.

The use of Mule led to the discovery of Mule partner U1 Technologies and their Ambrosia messaging platform.

And they're hiring.