Phil Windley's Technometria

« July 09, 2003 | Main | July 11, 2003 »

July 10, 2003

Mike Kruckenberg: Transforming XML for Web and Print

Mike Kruckenberg is from Tufts University. He's talking about how they built a system for managing documents and and displaying them for various media (i.e. content management). Mike, in case you're curious, the brother of Pete, a good friend.

Mike specifically concerned with translating documents for web and print (namely PDF). They created a document standard with a Schema and developed templates for XML authoring application to make creating the documents easy. They created an customized XML authoring environment from an off the shelf tool that was essentially the destination for any conversion process. They also provided an online tool for people who didn't have access to the authoring tool.

Existing HTML documents were cleaned up with Tidy and then a homegrown tool translated the cleaned-up HTML to XML. Once the XML was valid, the XML document was put into the database. For MS WOrd document, they tried a bunch of things, wvWare, saving as HTML, saving as RTF, and third party stand alone tools. They're looking forward to WordML. PowerPoint is a big tool for faculty, so it had to be easy to convert to an XML document. For PowerPoint, they have a service which create an XML document from the text and save JPEGs out and wraps everything up in XML.

Here's some questions about conversion I'd ask if there was time:

• Did you try reading Office documents into OpenOffice and then transforming the resulting XML?
• Did you try saving as PDF and then converting that to HTML?
• Are you supporting emerging standards like SlideML?

The transform is done using the libxml2 and libxslt libraries fro Gnome because the have good performance and command line and Perl interfaces. xmllint validates XML against a DTD. xsltproc renders XML as HTML.

Just rendering HTML isn't the goal however. The goal is to render HTML and PDF for print. Mike and his team used FOP and XSL:FO to create PDF.

Mike gives some lessons that they learned:

• Ensure XML is well formed and valid
• Lack of structure in the source document results in meaningless XML
• Special characters require the use of entity mappings
• Using the tool must be convenient
• FO transformations have limitations--read the documentation
• Fonts in PDF can be problematic and require embedding fonts
• Image and spacing issues cause problems and users don't understand the limitations
• The processes can be slow and CPU intensive so PDF documents need to be pre generated, not done in real time.

CIO Magazine published an article about this project.

06:49 PM | Recommend This | Print This

Brian Ingerson: Ingy on Kwiki

I'm in Brian Ingerson's talk on Kwiki. Kwiki is the PERL based Wiki software that is running the OSCON Wiki. Brian is the author of numerous Perl modules. One of the chief design goals behind Kwiki was to make it easy to install. Brian demos this by creating a new directory (that can function as a CGI directory), typing "install-kwiki" and there's an instant kwiki. That's a neat feature if you want to tack up and tear down wikis for specific purposes (like using them as an adjunct to a conference call).

Brian points out a few Kwiki sites: Quiltzilla and LondonGeek.org. There are others.

Brian does a demo showing how to overload classes to change the formatting. he does this to show how the show the object oriented design and the overall design. The code is nice and clean and the design simple. Overall, this looks like a good, easy-to-use Wiki tool. I used TWiki for my class this last spring and it worked, but was difficult to configure and set up. This fall, I'll try Kwiki.

04:05 PM | Recommend This | Print This

Panel: Open Source Projects in the US Government

Lisa Wolfisch is conducting a panel on Open Source Projects in the US Government. Well, its actually just her and Pat Moran from NASA Ames. The third panelist was supposed to be Terry Bollinger, but he couldn't make it at the last minute. That's too bad, Terry is a MITRE employee who did a study on FOSS (free and open source software) usage in the US government. I heard his speach last January and it was full of interesting things. Lisa said she has his slides, so maybe she'll give us a rundown.

She is going over a summary of Terry's information which showed 110 projects using FOSS in the DoD, with infrastructure and research projects being the most strongly represented. The DoD CIO placed FOSS under the same requirements us commercial software. There are, obviously, requirements for security certifications (like NIAP and Common Criteria). Oracle and IBM are sponsoring versions of Linux for Common Criteria evaluation.

EAMS, the Enterprise Architecture Management Software group, in the federal government is using an OSS model to support shared EA software.

Lisa is now discussing her project which is the State and County Quickfacts at the Census Bureau. The site features thematic maps designed for online viewing. The project was unfunded and took six months from planning to release. The site is built on a LAMP platform for $0 in start-up procurement. The same code drive MapStats which shows state and county profiles on FedStats.gov.

Lisa cites the fact that OSS has no procurement delays as a big factor in choosing it for government projects. Projects often die when there's a funding delay, even if the money shows up eventually. OSS0-based projects have an advantage in that area.

Looking around, the room is full with people sitting on the floor and standing at the back. Its a fairly large room too.

Pat wrote a paper in support of FOSS that includes quotes from the NASA mission statement about providing for the widest and appropriate dissemination of information. Some recent progress at NASA shows the legal office saying that there are no barriers to releasing software as open source from NASA. The next step is to work with the "Software Release Authority" within NASA to develop and OSS process.

03:06 PM | Recommend This | Print This

Doc Searls: DIY-IT: How Open Source is Turning IT into a DIY Marketplace

Doc is speaking on DIY-IT, his view of how OSS is turning IT into a do-it-yourself marketplace. This talk was added just today and I'm glad to see Doc on the program. He's always got something interesting to say and further, he says it in an interesting way. Doc's July column for Linux Journal is Linux for Suits: How Linux Makes Companies Smarter and I'm confident that's related to what he's going to say today.

Every story has three parts: a story is about a (1) character with a (2) problem moving toward (3) resolution. Doc says this is why sports is so popular. Good characters thicken their own plots (tell me about it). War metaphors are great for describing problems "MS Preps Assaults on Linux." Doc says that marketing people fail to tell a compelling story when they try to portray the company as perfect (i.e. no problems).

There are two stories about Linux in the enterprise: the outside story about what vendors are doing for the customer and the inside story about what customers are doing for themselves (may or may not involve vendors). The first story is about attractive executives doing battle for their customers. The second story is about poorly dressed geeks. Which is easier to tell?

Doc tells some inside stories. The first is about Roland Smith and LSI Logic. The second is about Leon Chism and Orbitz. The third is Greg Thompson and UCAR (University Corporation for Atmospheric Research). The fourth story is Elliot Noss and Tucows. The fifth story is about Paul Perry and Verizon (who as on my panel at the Weblog Business Strategy Conference). The sixth story is David Pippenge and Yarde Metals.

The outside view is simple: vendor gives goods and services to customer in exchange for money. There are plenty of stories that tell well in that context. The real world is more complex with developer communities surrounding all this that interact on both the vendor and customer side. The use value of IT in this context is much greater than the transactional value of IT in the simple view. Most of what happens in this context doesn't tell well with the usual story metaphors about sports and war. There are stories.

The software industry is still growing up. The software industry is maturing into something like the construction industry. "We work in crews on projects." Do-it-yourself (DIY) is at the heart of getting stuff done. OSS is making DIY-IT possible. Its how the demand side supplies itself without a vendor relationship. That doesn't make vendor relationships bad, but you don't always have to go outside to solve a problem. Commercial vendor tools also drive the DIY-IT.

The construction industry is the oldest industry and is worth $2 trillion worldwide. Sharing know-how is natural in the construction industry. Doc makes a joke about a construction worker claiming his way of hanging a door is protected IP to a big laugh. Commodities are okay in the construction industry and big companies make good money in those commodity businesses. There's room for everyone in the construction industry because people are always trying to get things done and you've got to build things to do that. These are all good metaphors for where the software industry is headed.

01:06 PM | Recommend This | Print This

Kevin Falcone: LDAP: Integrating Authentication Across Operating Systems and Applications

Kevin Falcone is speaking on LDAP: Integrating Authentication Across Operating Systems and Applications. This talk is a report on work he did as a student administrator at Northeastern University to create a single authentication authority using LDAP. This is a popular talk. There are people sitting on the floor and standing at the back and out the door.

The old system was based on NIS (Network Information Services or yp). There was no security model, the passwords were passed in the clear, and you have to use the yp tools. On the other hand it worked because its stock on Solaris and can be integrated with Linux, BSD, OS X, and even Windows.

Kevin's goal was to replace NIS with LDAP in order to increase security. LDAP can be used with SASL (Simple Authnticatino and Security Layer) or TLS using SSL certs. Moreover, the password file can be protected. The downside is that LDAP is complex, mirroring and replication are more difficult than in NIS, configuration is complex, and there are few tools for managing data. There's no standard for transfering data between different LDAP tools like OpenLDAP or iPlanet.

Kevin decided to use OpenLDAP even though he had access to a copies of iPlanet (educational institution) and the LDAP server in Lotus Notes (University standard). Of course, if he'd chosen one of those, he wouldn't be presenting here, would he? :-) He also used OpenSSL for the SSL layer (TLS). He chose TLS over SASL because of the simplicity of channel encryption. There were some systems that wouldn't support TLS, so he did use SASL (plaintext, Digest-MD5) in those cases.

There are predefined LDAP schemas for NIS which stores account data, password data, and access data. This, combined with the core schema (personal information) formed the basis for creating the data set. A set of scripts called PADL can create LDAP files from /etc/{passwd,group} and NIS data. It works great the first time, but doesn't go back and forth. Kevin ended up writing his own scripts using Net::LDAP.

Kevin created a testbed consisting of one Solaris 9 machine, multiple Debian Linux machines, an OS X machine, and multiple Windows machines. To make it work on Linux, Kevin used PAM. This talk generates some sigificant comments in geekspeak. Several comments of "did you try...." followed by some detailed discussion.

The Solaris LDAP client works with OpenLDAP and also uses PAM, similar to the way it works in Linux. The client manages the configuration files automatically, but its managing files that can be monitored outside the tool. OpenBSD doesn't come with an LDAP client out of the box, but there's one in the ports tree that works. There's no NSS, so you have to edit the /etc/passwd file to tell the machine that a user is an LDAP user. OS X was trivial. You tell it to connect to a particular LDAP server in the directory access panel and it "just works." Go figure. LDAP can be used with Windows by syncing with the Active Directory server, but leaves a problem of one way data. Again, go figure.

Conclusions:

• LDAP is difficult to configure and implement
• The gains in network security are significant
• OpenLDAP libraries and Net::LDAP work well for integrating one-off applications

12:07 PM | Recommend This | Print This

Mitch Kapor: Linux' Journey to the Mainstream Desktop

Mitch Kapor is talking on "Linux' Journey to the Mainstream Desktop." OSAF's larger mission embraces more than just Chandler. Mitch recently initiated a project on behalf of OSAF to "take a careful look at the state of Linux on the desktop, and asked Bart Decrem to spearhead a short-term research project to assess the current situation and trends." You can read that report here (PDF).

Mitch is convinced that Linux will take a significant share of the desktop market. He takes a swipes at SCO as a company that has no business model other than taking the money that other companies have earned through litigation. He didn't actually say "SCO" but everyone knew what he was talking about and applauded. This, he says, is a sign of success for Linux. He cites several trends:

• PC Commoditization
• Increasing trouble getting consumer and companies to invest in continued upgrade cycles.
• Increasing feelings that companies (Microsoft) are using exploitive licensing.

Mitch references massive deployments of Linux desktops, mostly with a public sector angle. The largest one is Thailand's decision to deploy 1 million low cost PCs inside the country.

Transactional workers, people who use computers to perform some specific task, are the next trend in Linux deployment. Call center workers are examples of transactional workers. Knowledge workers use more apps and are more flexible in what they do each day than transactional workers. Getting significant numbers of knowledge workers to use Linux will not happen until at least 2007. The total breadth of applications available under Linux doesn't suit their needs yet.

Mitch gives a report card for Linux on the desktop:

You should read the report (linked above) to understand the reasoning behind the grades. Mitch also gives a report card for the OS Desktop ecosystem. The bad grades (D) are in the areas of ISV's and distribution channels.

Much of the remaining work either spans multiple projects, or has fallen through the cracks between them. He provides a technical agenda which breaks down as 50% about office file formats, 30% about strengthening the foundations of the desktop, including a hardware abstraction layer and desktop consistency, and 20% about fit and finish. Remember this is an agenda for gaining wider adoption in the transactional worker market, not knowledge workers.

Predictions

• MS price cuts
• From good to gooderer
• 10% share of desktop globally running Linux in the not-to-distant future
• Rest of the world leads US as adopter
• Public sector is a driver in adoption
• Selective adoption in enterprises
• No consumer momentum for a while.

OSAF is doing things to attack the problems:

• Exploring ways to facilitate a desktop foundations layer
• Funding extensive compatibility tests with respect to Excel
• Providing fiscal agency services to selected OS projects
• Giving resources to public sector decision makers at opensector.org
• Building high quality desktop applications like Chandler.

11:11 AM | Recommend This | Print This

Stormy Peters: Open Source at HP

Stormy Peters is the Director of HP's Open Source office, the office that is responsible for HP's use of open source software. She claims $2 billion in Linux related revenue at HP last year. She has a nice slide that shows a hierarchical representation of open source licenses. She characterizes the MIT and W3C licenses as having no restrictions, the BSD and Apache license has having restrictions, but no impact on other code, and all the others as some variant of copyleft.

Stormy talks about Martin Fink's book called The Business and Economics of Linux and Open Source. Martin is the GM for HP's Linux group. I haven't read the book, but I may try to pick a copy up. The book jacket advertises it as a guide for business managers considering using OSS in their business. Disclosure: Martin is Stormy's boss.

She suggests the following business models around open source:

• Commercial software - Oracle running on Linux is the example she uses.
• Support and services - This is the professional services model.
• Aggregation and enhancing - This is Redhat and other Linux vendors.
• Commercialize with a dual license - "Free for non-commercial use."
• Enable hardware
• End of life - What to do with a dog product that isn't selling?
• Building an ecosystem - Eclipse is the example here.

Why would you want to open source a product?

• Commoditizes a market you don't control (disruption)
• Make a technology pervasive
• Promote a proprietary product you have
• Lower the overall cost of a project (shared effort)
• Promote hardware
• Enable custom solution for customers (let them roll their own)
• Exit a business
• Leverage resources from others

When isn't it appropriate? This is bound to be controversial?

• The product is a control point (Windows)
• The product is obsolete (Windows---NO she didn't really say that.)
• The cost doesn't justify the benefit. This is a nod to the fact that open source development isn't free.
• Misdirection and defocusing of resources
• Intellectual property risk cannot be justified. Don't open source something you can prove you have the right to. This is important.
• Don't open source something to compete against the OS community.
• Just because its cool (I disagree with this---this is a great reason to open source something---ofttimes you don't see the benefit until people play with it and geeks are the ones to do that).

She talks about why and why not to do OSS development in a company. The most interesting one, to me, was time to market. If certain features are critical for what you want to get out of the software and you can't control the release dates (its someone else's OS project) you may want to avoid OSS.

10:35 AM | Recommend This | Print This