Edited transcript of weekly BriefingsDirect[TM] SOA Insights Edition podcast, recorded April 27, 2007.
Listen to the podcast here. If you'd like to learn more about BriefingsDirect B2B informational podcasts, or to become a sponsor of this or other B2B podcasts, contact Interarbor Solutions at 603-528-2435.
Dana Gardner: Hello, and welcome to the latest BriefingsDirect, SOA Insights Edition, Vol. 17, a weekly discussion and dissection of Services Oriented Architecture (SOA) related news and events with a panel of industry analysts and guests.
I'm your host and moderator, Dana Gardner, principal analyst at Interarbor Solutions, ZDNet blogger and Redmond Developer News Magazine columnist. Our panel this week, and this is the week of April 23rd, 2007, consists of Joe McKendrick.
Joe McKendrick: Pleasure to be here, Dana.
Jim Kobielus: Thanks a lot Dana. Hi everybody.
Dave Linthicum: Thank you very much, Dana.
Todd Biske: Thanks for having me again, Dana.
Second, and in a possibly related area, we’re going to discuss Red Hat's acquisition of MetaMatrix. There’s also some other news by Red Hat this week about some bundling and a concentration on SOA and it’s values.
Last, or perhaps third, I saw an interesting post by Joe McKendrick this week, something around IBM dissing UDDI and calling for new standards. We’re going to ask him about that. And also, Dave Linthicum had some posts this week in his blog about a Gartner study and the good news and bad news about SOA. We’re going to ask him to fill out that topic a little bit.
But, let’s dig into the first issue of the day. This was brought up by Jim Kobielus, the whole notion of Semantic Web. It’s been perking up on the radar and showing up in a lot of different places. We wanted to see if there’s anything new going on and if there’s any convergence with what we refer to as Web 2.0 and even Enterprise 2.0. So, let me first go to Jim Kobielus. Jim, why are you interested in the Semantic Web and how would you define it?
Kobielus: Well, Dana, I'm like everybody else on this call. I've got a very reptilian, scaly, analyst skin, and so I'm a bit cynical about the notion of a Semantic Web, as many are, but nonetheless, the topic keeps coming back to the surface. I saw a couple of articles in the past two weeks, in which the topic of the Semantic Web got trucked out again. This time, I saw one article where it was called Web 3.0, and I thought, “Oh, my Lord, we haven’t even decided that we are all in agreement on the notion of Web 2.0.”
So Semantic Web, as a concept, is still in play, and as you are probably all aware, the term was essentially coined or popularized by Sir Tim Berners-Lee, the inventor of the World Wide Web. There is an activity at the World Wide Web Consortium that’s been going on for a few years now to define various underlying standards and specifications, things like OWL and Sparql and the whole RDF and Ontologies, and so forth.
Kobielus: Right. It’s still in play as a concept, and people are still energized to invest it with meaning and substance. So, what is the Semantic Web? Well, to a great degree, it refers to some super-magical metadata description and policy layer that can somehow enable universal interoperability on a machine-to-machine basis, etc. It more or less makes the meanings manifest throughout the Web through some self-description capability.
It’s a "boil the ocean" type of initiative, and it’s been going on for a few years. Of course, when you boil the ocean, you generally get a lot of plankton. So, is there anything useful in this plankton that’s been boiled out of the ocean of Semantic Web? I think there is something, and I am not going to bring in the whole notion of Web 3.0. It’s a distraction, unless you’re going to tie it right back to this week’s news.
As you mentioned, one thing that happened this week was that Red Hat, the open-source company, announced its intention to acquire a company called MetaMatrix. MetaMatrix is in a niche of the data management arena that’s often called either EII (enterprise information integration) or data federation. They're not the only player in this space. There are many others, IBM, Business Objects, Ipedo, Sybase, Composite Software. Many of these have EII products and what they all do is provide a semantic layer that essentially virtualizes and abstracts heterogeneous data resources, mediates data dynamically on the fly, and presents that data in a unified data model or structure to end users.
Kobielus: Yeah, primarily it’s being done within enterprises, but to varying degrees it’s being done across B2B environments as well, where you have diverse data on your ERP system. You have your line-of-business apps, transactional databases, web sources, and so on, and you’re not consolidating that data into a data warehouse. Rather, you're leaving all that data in the source repositories and dynamically mapping it to composite, unified views. What I am getting at is that there are a lot of vendors of these EII products that provide a semantic layer. Probably, the most widely used is Business Objects. They call it "Universe" capability. So, to some degree, this sounds like the Semantic Web.
Kobielus: The World Wide Web, as a phenomenon, really took off when it was adopted inside companies in their intranets. So, the whole notion of a "semantic Web," to the extent that we can all agree on a definition, won’t really come to the fore, until there is substantial deployment inside of enterprises. I agree with you there.
Conceivably, these EII vendors are providing a core piece of infrastructure that could be used to realize this notion of a Semantic Web, a way of harmonizing and providing a logical unified view of heterogeneous data sources. Red Hat, one of the leading open source players, is very geared to SOA and building an SOA suite. Now, they are acquiring an EII vendor, which itself is very SOA focused. So, you’ve got SOA; you’ve got open source; you’ve got this notion of a semantic layer, and so forth. To me, it’s like, you’ve stirred it all together in the broth here.
That sounds like the beginnings of a Semantic Web that conceivably could be universal or “unversalizable,” because as I said, it’s open source first and foremost. The super-distribution model of open source, a la Red Hat, will conceivably propel some sort of Semantic Web into universality. I think this is an important event. Anyway, I want to hear what everybody else on the panel thinks about this notion.
Linthicum: I'm a little confused about why it has taken us so long to realize this, because it’s been around for such a long time. I was just doing some Web searches, and I wrote some pieces three or four years ago around the Semantic Web and it’s use and play in both integration and SOA. I was kind of taken aback at the time that no one really cared, because I thought it was a very valuable notion. If we build on this, it does solve a lot of key problems. You end up dealing with universal semantics, how that relates to B2B domains, and how that relates to the enterprise domains.
As I'm deploying and building SOAs out there in my client base, semantic mediation ultimately is a key problem we’re looking to solve. This is the point you just made Dana. We have to have a better grasp on the data, how to extract that data into data services, and then apply those data services into our SOAs.
Linthicum: I think so, and I think there are a couple of opportunities here. We like to test things out on the Web and then drive it into the enterprise. There are a number of instances of that. Even the AJAX deployments I see today are more enterprise based than Internet delivered based, and the same thing will happen with the Semantic Web. It’s a key, universal standard that lots of people are accepting. People are going to vote through the usage of it, ultimately figure out if it's going to have value on the Web as an inter-company thing or Internet universe thing, and then drive that into the private SOAs that people are building.
Biske: I’m going to take my usual, pragmatic approach with this. First off, I agree with Dave that semantics is a really a critical piece in being successful with SOA. And I challenge the notion that you’re going to run into it whenever you’ve got these data-integration scenarios, whether it’s B2B or completely internal to the company. The issue that we run into is that it's still a very specialized discipline. There are a number of people, and I am not one of them, who really understand the technologies behind it deeply, and are continuing to push forward, but they haven’t yet found a way to really have it make sense to the average developer.
The average developer is still focused on the functionality of the business solution that they're providing. They know that they may have data in two different formats and they view it in a point-to-point fashion. They do what they have to do to make it work, and then go back to focusing on the functionality, not really seeing the broader semantic issues that come up when you take that approach.
It’s no different than just building out an SOA. People are still thinking in terms of their silos, and not thinking outside of the box and how it’s going to impact things in a long-term fashion. So, until we are able to have this hit home to the developer community, and really have them understand how they can take semantics, they can take that metadata, incorporate it into their development processes. It’s still going to be a very slow go.
McKendrick: It’s going to come from both directions -- from the top of the organization and from the developers within organizations. How’s that for a definitive answer? On the part of corporations, there is always resistance to exposing too much data and exposing too many of their applications to the outside. A little bit of this has been melting away, as we’re seeing the rise of software as a service, and from surveys and anecdotal data.
Information I’ve seen in recent months indicates that there is lot of interest in the whole software-as-a-service model. A colleague of ours over at ZDNet, Dion Hinchcliffe, encapsulated the view of Web 2.0. He calls the Web 2.0 "The Global SOA," meaning that a lot of the standards and services that you see in an SOA behind the firewall these days will be available in a global sense from the Semantic Web or from the Internet at large. And a lot of the services, standards, etc. that we will adopt and see in companies will gradually, more and more come in from the outside -- from the Global SOA, if you want to call it that.
Kobielus: Are you asking whether EII and MDM are different approaches? Can you clarify?
Kobielus: That’s quite right. Basically, you can look at semantic interoperability as being the global oceanic concern. Wouldn’t be great if every single application, data base, or file that was ever posted by anybody anywhere on the Internet somehow, magically is able to declare its full structure, behavior, and expectations?
That’s the "boil the ocean" aspect of it that everybody keeps focusing on, but then you can look at semantic interoperability in a well-contained way as being specific to a particular application environment within an intranet or within a B2B environment. Many vendors in the master data management or EII space are approaching it from that direction. In other words, they are providing pre-built domain models with standard data definitions, metadata, ETL rules, governance workflow rules, etc., as a standard component of their applications.
These are applications running out like an MDM or an EII substrate that they will sell in to particular niches or segments, such as enabling customer data integration within an organization or enabling a supply chain optimization on a B2B basis, where there is essentially this shrink-wrapped semantic model that’s provided as part of the overall application. That is semantic interoperability, because what’s happening then is that the vendor, and/or its ISV partners, and/or the customers are taking the semantic domain model and modifying it to their particular requirements.
So, that is very much, in a sense, the Semantic Web. It’s semantic interoperability out of the box that doesn’t rely necessarily, or very much, on things like RDF and all of these other specifications that are sitting on the sideline and are represented in today’s EII market.
Biske: It’s interesting that you bring that up, because last week David Margulius published on InfoWorld an article titled “Corporate librarian replaced by Web app.” He wound up having a link to one Clay Shirky article that talked about removing the shelves. He was comparing it to a library and the whole Dewey Decimal System and saying that what Google did was to pull all the shelves out. That is to say that there is no one way of organizing it, and it's really focused on the relationships. So, even there, I don’t know that Google’s necessarily created a Semantic Web. They have created a relationship index.
Biske: So, you’re still not going to get that semantic information out of it, but it's probably a far more appropriate way of getting at the information, because, as we know, there are so many different ways of slicing and dicing that it’s a challenge. It comes back to why some of the semantic technologies haven't made advances. The challenge within data federation is that we keep trying to find one universal model, when there isn’t one.
Biske: Where a Google approach can help, and where Semantic Web comes into play is when we know there's a relationship between these two documents, two things are accessible via URI. We need a standard way of getting metadata about that, beyond just the tags, because we found that while the tagging is a step better, you still have a semantic issue with the tags themselves.
Now, if we add in the metadata that provides additional context about the relationship, what is that going to enable? Now we can start to really make semantic connections between these things and start to leverage that metadata in an automated fashion.
Linthicum: I think so. I’m just going to add what Todd said. One thing that’s going to happen with the influence of something like Google, which is having a ton of a push in the business right now, is that ultimately these guys are exposing API’s services. They just came out with a streaming API last week, and its actually pretty good. But, at the end of the day, it’s a service that you invoke to get a stream out of it. They're coming to the realization that the developers that leverage these APIs need to have a shared semantic understanding out on the Web. Once that starts to emerge, you're going to see a push down on the enterprise, if that becomes the de-facto standard that Google is driving.
In fact, they may be in a unique position to create the first semantic clearing house for all these APIs and applications that are out there, and they are certainly willing to participate in that, as long as they can get the hits, and, therefore, get the advertising revenue that’s driving the model. But, I think the indices and the ability to catalogue not only content, but services and then information around services -- and not only services, but data and information around data, metadata -- is really the next logical step in that area. There is going to be a big push there in the next few years. I don’t necessarily see it as the 2007 opportunity, but it’s clearly a 2008 opportunity.
Gardner: That’s very interesting, and I don’t think I've seen this declared elsewhere -- Google, being in this unique position, perhaps more than the W3C or Oasis or some of these other organizations, could create some sort of an approach for de-facto industry standardization around indices and then, ultimately, Semantic Web types of activities.
Linthicum: If you think about it, Dana, Google right now is probably the number one provider of services out on the Internet, providing the Google maps, API’s and everything else that’s out there today, and they see this as a strategic direction where they can grow. So, they are in the API business and they are in the services business. When you're in for a penny, you're in for a pound, and you get into that world. So, you start providing access to services, and rudimentary on-demand governance systems to account for the services and test for rogue services, and all those sorts of things. Then you ultimately get into semantics, security, and lots of other different areas they probably didn’t anticipate that they'd get into, but will be pushed into, based on the model they are moving into.
Kobielus: I agree with Dave. Google potentially plays a very important role in catalyzing the spread of something like a Semantic Web, simply because they are the master indexers of the universe. When you look at the EII products and the vendors that are out there, every single one of their semantic layers, at its very core, has an index. It’s a registry of the data and the metadata that exists out in whatever universe or other environment that that particular vendor’s product is managing.
So, to the extent that Google is the default index that everybody goes to first and foremost, and Google crawls and has indexed and documented every visible piece of content in the universe, I think they sit right dead-center of the heart of the Semantic Web, or will, or somebody like them, somebody who plays that role.
Biske: I agree with those comments as well. If we take this inside the enterprise, and the enterprises deploy the Google appliance, they can go and crawl all of those sources of metadata. The enterprise is trying to take on some form of data governance and come up with their canonical models, and they need to take a bottom-up approach to all of the databases that they have got. Why not throw your Google appliance at it and then start throwing in the keywords that the data-governance committee is actually debating on how to describe the data? You’ll instantly have all the relationships between not just what's in your relational databases, but your unstructured content as well.
They really do play a key role, even indirectly, in the SOA space. One of the recommendations I always make to the companies that I’ve worked with, if they are looking at register repository, is to ask the vendors whether their repository can be indexed by a Google appliance. Ultimately, that’s going be how the developers are going to find the services, and they want to use Google to find it.
Kobielus: Here is the thing. For this grand notion of a Semantic Web to be truly universal, all the many indexes and search engines that are out there need to be federated. Semantic interoperability is a huge problem that I can scarcely put my head around. How is that going to happen?
If you think about it, the analogy is a whole federated identity world that’s been percolating since the start of this decade. You have concepts or standards like SAML, the Liberty Alliance Standards like IDFF, and now SAML 2.0, WS Federation, etc. Those guys have not converged on a common, universal federation environment, in which everybody participates. And, that's just in identity management. Think about federating metadata repositories across the known universe. That should be a tough nut.
Linthicum: I don’t think that they have gone that far-far down the road.
Linthicum: I think we are giving free strategies to a multi-billion-dollar organization. I don't know if you saw the release this week from salesforce.com. They are in the platform business, so they deliver semantic mediation, app development and integration, and even aspects of SOA on demand now. So, that’s another player in this game, as well, and a much more sophisticated player, when you consider that they give you the coding tools, design tools, the database, and the platform in a virtual operating system. There's going to be a lot of people moving at that platform, and they're building SOAs on that platform and extending them in their enterprise. They could become the platform for semantic mediation as well. That’s their opportunity right now.
Linthicum: I think we are all waiting for checks from Microsoft and they haven’t come yet.
Kobielus: I was thinking that the real platform for semantic mediation are these industry groups that define common data models for those industries. They can form those data models that then are the platforms for semantic interoperability within those environments and within those industries. So, that’s also important. Doesn't OMG under Richard Soley develop, to some degree, those data definitions on a industry-by-industry bases or catalyze their development?
Kobielus: Because it’s all about taxonomy governance, and governance quite often is industry specific.
Biske: I actually do have a Microsoft comment. The one thing that I will give Microsoft credit for is that of any of the players, they are probably one of the ones most likely to bring metadata into the developer’s world. One of the WS-Star specs that I like a lot, but that you don’t hear too much about, is the WS-Metadata Exchange. Microsoft’s got an implementation of it, and they are strongly trying to leverage that in the advancements they are doing to .NET and Windows Communication Foundation. Because of how developer focused they are, if they continue to push that, they can start to bring some of the Semantic metadata into the developers' context and show them how we can leverage not just in design time but in run time. So, we’ll just have to wait and see what they come up with
Biske: I don’t know that we’ll necessarily see that, because Microsoft's strength is in development tools, whereas Google is more in providing their widgets that can be combined on the Web pages and the whole
Biske: That will be the challenge, if Google actually tries to come up with a standard for accessing them, and Microsoft wants to push one of their own. But I haven’t heard too much on either side about doing that.
McKendrick: Sure. Actually, IBM released its Websphere repository and registry product's latest release the other week.
McKendrick: Exactly, and Burton Group's Anne Thomas Manes released a report in which she pointed out that there really isn't strong support for UDDI. The product supports UDDI implementations outside of their own registry, but its more of a proprietary solution, kind of a step backwards. IBM was one of the co-creators of UDDI in the first place 8 to 10 years ago, along with Microsoft and Ariba.
It was originally created as kind of an e-commerce yellow pages. There would be this place where you could go to find out about services, and it would link the services across the Web or Internet. Of course, it didn’t really happen. In fact, these companies maintained public UDDI registries, and they shut them down because it just wasn’t getting them traffic and they couldn’t keep them updated. But UDDI found a new life as the internal registries for SOA governance.
In an accompanying article that responded to Anne Thomas Manes' observations, IBM came out and said "Yes. We are not exactly robustly supporting UDDI any longer. We feel that UDDI does not support the vision of SOA as it exists these days." It supported Web services, the interchange idea of Web services, at an earlier era, but SOA is an internal building block function, and UDDI is not robust enough.
McKendrick: They say that in its current state, the UDDI standard doesn't support the lifecycle aspect of service development. Again, it’s a more than an exchange and communication vehicle, but it does not have that from design-time-to-run-time support that registries should have.
They're trying to do just that, create a lifecycle approach to services, so that developers who collaborate with one another, perhaps across global distances and multiple organizations, can be in a common mentality around the governance and the use of these services. They can relate to operational governance issues fairly early in development. Has anyone else been made aware of Jazz or looked into that at all?
McKendrick: Well, IBM said that the vendors should go ahead and start working on a new replacement or a new generation of registry, and maybe this is what they are talking about.
Biske: It's unfortunate that the IBM article really was very lacking in specifics. They said UDDI didn’t cut it, but they didn’t give any details on where it didn’t cut it or what advanced features their customers were asking for and needed. It doesn’t surprise me that there are some ramblings around this. One of the things that I have called out a couple of times my blog is whether we're willing to see convergence around all these metadata repository spaces.
I usually point to configuration management databases versus the UDDI and registry repository space. Clearly, it's not a great leap to get into the policy management space, which now hits the heart of governance and the topics around SOA registry repositories. So, there are some clear ties there. If each one of these systems has its own unique way of representing the metadata, as well as accessing the metadata, then perhaps IBM is right that we do need more work on the standardization front to figure out what is the best way for just accessing this universe of metadata, whether its policies for the run-time infrastructure, additional semantic information for design time, or both.
Linthicum: I think that’s a response from Laszlo and a bunch of other folks in the
Linthicum: Yeah, open source, discounted, and therefore making it much more attractive. They add some value, but in many instances, it's just becoming a gimmick to sell software, and this may be an instance of that.
Linthicum: Absolutely. I have a lot of experience with Flex, as I built Bridgeworks, a company I was CEO of a year ago, around Flex, and it’s a very good tool, and it does a very good job. I think that they are trying to get it out into the development community to get away from the whole Flash Internet content, Internet sex appeal delivery industry they’re in right now and move into the app development industry. That’s a great way to do it, because people say, "Open source? Okay. I'm going to give it a shot."
Linthicum: That’s a tough leap for them, because ultimately that’s their bread and butter. They own a lot of that market, and they do so with a very well-thought-of product. They are probably not looking to do anything different, because that’s going so well. The Flex market is completely different. You've got a lot of these folks entering that market with some very nice technology, and I think Adobe is just responding to that competitiveness, versus owning the market. They have to compete in the market, and that’s the way they are going to sell their product. I think it’s a smart move for them.
Kobielus: I'm sure that’s their hope, that Flex can catch fire with the development community by being freely available everywhere. That’s the same gamble that Red Hat took with MetaMatrix, when they open sourced that, but it doesn’t mean that they are going to succeed. There are lots of alternatives to MetaMatrix in the EII space, and I am sure that there will more of that coming along. So, it’s definitely a gamble for Adobe to give it away.
Linthicum: The take away on that are things I've been seeing for quite some time. People are actually getting into the project-level of SOA now. I am seeing that happen a lot. I am sure MomentumSI and other folks in the service-delivery area doing that, and they're finding out that it’s hard. It’s tough to do, and it’s a systemic change in the way they think about architecture. There’s a bit of a hangover from the hype that occurred in the last few years. So, people are having a bit of a negative take-back on that. I feel like I'm the designated buzz-kill running around the country, working with my clients, introducing them to the fact this is not something that can be taken lightly. You can’t bolt it on. SOA is something you do, not something you buy.
So that’s the bad news. The good news that Gartner found out is that the concepts are working. People are successful with the technology, and I'm seeing that in my practice as well. It’s a valuable thing, and they can make a lot of money in terms of expense savings and the strategic advantage of agility that goes right to the bottom line.
So it’s a good concept. Conceptually it’s working, people are proving it out in the market. It’s just a bit different from the way in which the vendor hype has been describing it over the last couple of years. It’s more of an evolution of hard things that need to be done. You can’t buy stuff out there. You can't buy SOA.
Biske: I actually started my blog before I joined Momentum so it’s under my personal domain. It’s www.biske.com/blog
This is Dana Gardner, principal analyst at Interarbor Solutions. I want to thank Joe McKendrick, Jim Kobielus, Dave Linthicum and Todd Biske for joining in yet another BriefingsDirect SOA Insights edition podcast. Thanks all for listening.
Listen to the podcast here. Produced as a courtesy of Interarbor Solutions: analysis, consulting and rich new-media content production. If any of our listeners are interested in learning more about BriefingsDirect B2B informational podcasts or to become a sponsor of this or other B2B podcasts, please fill free to contact Interarbor Solutions at 603-528-2435.
Transcript of Dana Gardner’s BriefingsDirect SOA Insights Edition, Vol. 17. Copyright Interarbor Solutions, LLC, 2005-2007. All rights reserved.