MES 2006 - Port 25: The Open Source Community at Microsoft

MES 2006 by admin on May 02, 2006 12:08PM

Swimming is my cure for jet lag. I am currently at the MES 2006 (Microsoft Executive Summit) in Mumbai (Bombay), India – an annual event for the top 250 CIO’s in India. I’ve been here a couple days and have been waking up at 3am, so my cure has been a pre-dawn swim in the hotel pool. The hotel I’m staying at has a nice pool, right next to Powai lake and it’s protected forest, where I later learn is a natural home to leopards and alligators, which keeps me alert while swimming solo.

MES 2006 is a great opportunity to meet many of the top CIOs, IT decision makers and partners in India and I’ve been enjoying it immensely. I did a presentation on our platform strategy, people-ready businesses, and in particular how we think about ‘coopetition’ in this strategy. One of my favorite things about this job is talking to customers about their IT environments, issues and dreams. What has been fascinating about my conversations here in Mumbai has been the many different ways customers have designed and architected for interoperability. From banks to manufacturing companies to consulting services, almost every customer or partner I’ve met with has an interesting story about interoperability in their IT environment. No surprise really, heterogeneity is part of any large IT system, but the recurring theme I’ve noticed here is the pragmatism and clarity of focus on where and when interoperability is needed. And if you’ve ever spent time on the roads in Mumbai, you’ll realize that interoperability is part of everyday life!

One Microsoft partner I had dinner with explained a very large, multi-tier system they build and sell which uses Windows, Unix systems, and a mainframe – all for one application (it is a large and critical application, so this isn’t really overkill for what they do). Although they want to eventually migrate from the mainframe for cost reasons, they have chosen technologies to get the job done as best suited their needs and skills. And – importantly – they factor interoperability into every architectural plan, RFP, or design that they think about – it’s as important to them as feature functionality or testing. It’s a core part of their maturity model. So what do they look for to qualify something as ‘interoperable’? Open and mature standards that have industry wide acceptance. They also understand the difference between open standards and open source, and gave me a very lucid walkthrough of the differences. Simply put, they explained that open source is a development and distribution model and open standards are specifications that can be applied to interfaces and technologies to enable data exchange. It is that clarity that, I believe, has helped them to design for interoperability with their eyes wide open.

One more thing. I had a customer meeting where I heard a great description of IT value. We were talking about software utilization, the dreaded ‘application backlog’ that many CIO’s face (CIO magazine has a great column on this here). When I asked about their deployment experiences with Microsoft software, the customer told me a story about their instant messaging deployment. Within 48 hours of deploying Microsoft Live Communications Server for instant messaging and collaboration, they had over 16,000 people utilizing the product. He then said, “Listen Bill, it is actually quite simple, when I can deploy software that immediately 16,000 people start using on their own because it’s important and useful to them, that is value.” Clear and simple definition of value: people use it on their own volition - something we should all remember.

Although swimming does help jet lag, coffee is equally important, so it’s now time to go find some. Until next time, may you avoid stray leopards. -Bill

Email Digg It! Del.icio.us SlashDot It!

Comments RSS

davidmeyer said:

Hi Bill,

I'd be curious to know what the breakdown is with regards to operating systems. I know from the work I have done with clients in India and with clients who have Indian interests, Linux is a big player there. I'd be curious to know how they are making their environments talk. Are they using Samba or some other utility like that?

Enjoy India!

Dave

posted at 05:02PM 05/02/2006
billhilf said:

Hey Dave, from the folks I talked to, there is a mix of Linux (largely non-commercial), Windows servers and commercial Unix and some mainframe (os390, z/os). I was most surprised by the large number of Unix still in use (Solaris, HPUX, some AIX). The customers I talked to are trying to standardize at higher levels of interoperability, largely Web services, versus grunting over RPC or SMB related interoperability. This is, at least, the folks I talked with, but I think it's a sound technical strategy. There is also a fair use of enterprise service bus systems to broker messages and processes between disparate systems (Microsoft BizTalk, IBM WebSphere Message Broker, etc.). -Bill

posted at 06:41PM 05/02/2006
fluke said:

"So what do they look for to qualify something as ‘interoperable’? Open and mature standards that have industry wide acceptance."

There is several places where something is a open and mature standard that have industry wide acceptance but is not what I call interoperable. If you develope a new embedded device with web browser, one of the items that most users of the web come to expect of a web browser is support for Macromedia Flash. It is an open standard since you can download the full format description from Macromedia's web site. And is it a mature standard that has gained industry wide acceptance. But it is such a bloated standard with skewed differences between the documented format and what the offical implimentation provides that success of being interoperable depends on Macromedia's support.

Another problematic "web" format is ActiveX. MSDN has enough documentation that it can be considered open. And it clearly has become a mature standard with industry wide acceptance. But even when offered the source code to WinCE, supporting ActiveX on a new processor is prohibitive. The MSDN defined interfaces and technologies of ActiveX still does not en able data exchange to the new platform. Yet Microsoft has stated several times that it sees the importance of the web being a set of interoperable standards.

"We were talking about software utilization, the dreaded ‘application backlog’ that many CIO’s face"

One of the problems with "software utilization" is the lack of vendor backing in full utilization of software. Take for example Windows 3.x Macro Recorder (hopefully this is a historic enough example not to get labeled as MS bashing). My expierence was that users that attempted to "fully" utilize Windows 3.x had a harder time with migration. To get a solution with vendor support while upgrading Windows had the unexpected cost of having to purchase a third party replacements for the "full" features. There several cases where buying two seprate packages where only vendor supported core features where used in each would have been easier (and maybe even cheaper in the long run) than trying to accomplish everything by using features that later got labelled deprecated and eliminated. The users that "fully" utilize software tend to slow down the migration to software upgrades and require additional time/money to move their habits to a solution that is similar enough to gain acceptance.

"they explained that open source is a development and distribution model"

This could be read to imply that FOSS is simply a model by which implimentations is distributed. Yet FOSS also distributes the development model such that deprication of a feature is no longer up to a single vendor. Any other group with enough incentive can continue to promote development of features without having to re-invent the wheel.

posted at 07:44PM 05/02/2006
davidmeyer said:

"Hey Dave, from the folks I talked to, there is a mix of Linux (largely non-commercial), Windows servers and commercial Unix and some mainframe (os390, z/os). I was most surprised by the large number of Unix still in use (Solaris, HPUX, some AIX). The customers I talked to are trying to standardize at higher levels of interoperability, largely Web services, versus grunting over RPC or SMB related interoperability."

Hi Bill,

Fascinating. I too am surprised about the mix of operating systems. When you have time, I'd be curious to know what kind of challenges this puts on Microsoft. What I mean by that is this: It you had a customer who used all MS and Red Hat or MS and Solaris, that makes the job of those charged with system integration fairly easy. However, when you have a mixture of MS, HP-UX, Solaris, AIX, Mainframe, AS400, OpenVMS, etc. this can get tricky. I am wondering if there might be an opportunity there for Microsoft to create a "Translator" that can allo all of these boxes to talk to each other? Just a thought, because while they may not continue to spend on *nix, it probably isn't going away any time soon. Maybe this is a good Open Source project for the lab?

Dave

posted at 09:51AM 05/03/2006
einhverfr said:

Hi Dave,

I am not sure there can ever be a universal translator. New technologies come out all the time, so your universal translator is going to perpetually lock people into old and likely insecure technology. In essence, you would have a huge maintenance expense, and not a lot of payback.

The answer in my mind is to focus on standards and standards compliance. One can then offer a few products to integrate other infrastructures into your environment (SNA server, for example). Or better yet-- focus on one group of standards (Kerberos, LDAP, IETF standards) and add products to integrate other architectures (SNA, OSI protocols such as H.323, etc) into this architecture using gateway products and open source projects.

Finally, there will be places where heterogeneity of standards is not only unavoidable but desirable. FOr example, although there is a trend towards considering VOIP as an alternative to traditional circuit switched telephone service, there are good reasons not to run VOIP for mission critical voice operations (imagine if MS was using VOIP for their support call centers and you had a computer virus that generated enough internet traffic to disrupt inbound calls). Providing gateways between these different types of networks is another area of translation.

In essence, the best approach is not the universal translator but a bunch of little translators which all talk the same language. THey need not all come from the same vendor either.

Best Wishes,
Chris Travers
Metatron Technology Consulting

posted at 01:06PM 05/03/2006
jdzions said:

Microsoft already has a couple of "translater" products: MIIS, for directories, and BizTalk for application data flow. We also have a slew of "point to point" interop tools: SFU, Services for Netware, Services for Mac, Host Integration Server (mainframe), etc.

The problem with translation is that you either select a "least common subset" thru which all translation passes, or a "most inclusive superset". In the first case, any feature which isn't common to all translated worlds is simply lost in translation. In the second case, you have to build a tremendously complex metaverse to support all possible features, some of which may conflict with each other. Either way, someone's not happy.

posted at 06:18PM 05/05/2006
Wesley Parish said:

"The problem with translation is that you either select a "least common subset" thru which all translation passes, or a "most inclusive superset". In the first case, any feature which isn't common to all translated worlds is simply lost in translation. In the second case, you have to build a tremendously complex metaverse to support all possible features, some of which may conflict with each other."

The thought that occurs to me is that if that is the only option/s you give yourself, you are missing out on something. I would have thought the better option was to analyze the problem/s and work out which ones constitute a common set and which are external.

Then work out which of the common set constitute the core of the commonality and then, get them fixed up and working. Then you add on the modules that will add the most value most quickly, and provide the others when the customers start asking for it.

In other words, you start off from the "most common subset", get that tried and tested and working, then you can build on the other subsets until you have a working "most inclusive superset", which isn't going to bother anybody because they can take from it whatever they want and not worry about what they don't want.

At least, that's the way that makes most sense to me. I could be wrong, of course.

posted at 09:01AM 05/06/2006