These are my notes from the Advanced Web Development: software strategies for online applications workshop with Rob Stein, Charles Moad and Edward Bachta from the Indianapolis Museum of Art at Museums and the Web 2008 (MW2008) in Montreal. I don't know if they'll be useful for anyone else, but if you have any questions about my notes, let me know.
They had their slides online before the presentation, which was really helpful. [More of this sort of thing, please! Though I wish there was a way to view thumbnails of slides on slideshare so you can skip to particular slides.]
The workshop covered a lot of ground, and they did a pretty good job of pitching it at different levels of geekdom. Some of my notes will seem self-evident to different types of geeks or non-geeks but I've tried to include most of what they covered. I've put some of my own comments in [square brackets].
They started with the difference between web pages and web applications, and pointed out that people have been building applications for 30 years so build on existing stuff.
Last year's talk was about 'web 2.0' and the foundations of building solid software applications but since then APIs/SDKs have taken off. Developers should pick pieces that already work rather than building from the bottom up. The craft lies in knowing how to choose the components and how to integrate them.
There are still reasons to consider building your own APIs e.g. if you have unique information others are unlikely to support adequately, if you care about security of data, if you want to control the distribution of information, or if a guarantee of service is important (e.g. if vendors disappear).
Building APIs
They're using model driven development, using xmlschema or database as your model.
Object relational mappers provide object-oriented access to a database. Data model changes are picked up automatically and they're generally database-agnostic so you can swap out the back end. Object relational mappers include Ruby, Hibernate (also in .Net), Propel and SQLAlchemy.
IMA use Hibernate with EMu (their collections management system) and Propel. They've built an 'adaptive layer' for their collection that glues it all together.
Slide on Eclipse: 'rich client platform', not just an IDE. Supports nearly every language except .Net; is cross-platform.
Search
Use full-text indexes for good search functionality. They suggest Lucene (from apache.org) or Google gears. Lucene query types offer finer control than Google e.g. fielded searching [a huge draw for specialist collections searches], date range searching, sorting by any field, multiple index searching with merged results. Fast, low memory usage, extensible. Tools built on Lucene include Nutch (web crawler) and Solr – REST and SOAP API.
Bite size web components and suggestions for a web application toolkit
Harking back to the 'find good components' thing. Leverage someone else's work, and reduce dev/debugging costs – in their experience it produces fewer errors than writing their own stuff.
Storage – Amazon, Nirvanix, XDrive, Google, Box.net. Use Amazon S3 if accessed infrequently cos of free structure.
Video – YouTube, Revver, blip.tv also have developer interfaces. The IMA don't host any video on their website, it's all on YouTube.
Images – Flickr, Picasa. [But the picasa UI sucks so please don't inflict that on your users!]. Flickr support for REST, SOAP, JSON.
Compute (EC, Amazon web service) – Linux virtual machines. Custom disk images for specific requirements. Billable on use. See slides on costs for web hosting.
Authentication services – OpenID, OAuth.
Social computing
Consider social computing when developing your web applications – it's evolving rapidly and is uncertain. Facebook vs OpenSocial (might be the question today, but tomorrow?). Stick with the eyeballs and be ready to change. [Though the problem for museums thinking about social software applications remains – by the time most museums go through approval processes to get onto Facebook it'll be dead in the water. Another reason to have good programmers on staff and include content resources in online programs, so that teams can be more flexible while still working within the overall online strategy of their organisation.]
Developing on Facebook
Facebook API – REST-based API. Use their developer platform – simpler than original API calls. JSON simpler than XML responses. Facebook Query Language (FQL) reduces calls to API. Facebook Markup Language (FBML). HTML + Facebook specific features, inc security controls and interfaces features. [There's a pronoun tag with built-in 'they' if not sure of gender of person. Cute.] Lots more in their slides.
Widget frameworks
Widgets are the buzzword that hasn't quite taken off. The utility isn't quite there yet, so what are they used for? Players are Google, Netvibes (supports more platforms including Apple Mac dashboard, Yahoo, iGoogle, etc) but is Adobe AIR the widget killer? Flash-based runtime for desktop apps. e.g. twhirl. Run as background processes, and can access desktop files directly, clipboard, drag and drop. [I downloaded the AIR Google Analytics application during the session, it's a good example.]
Content management
The CMS is the container to put all the components together. A good CMS will let you integrate components into a new site with a minimum of effort. [Wouldn't that be nice?] Examples include Joomla, WordPress, Drupal, Plone.
There aren't slides for the next 'CMS tour' bit, but they gave some great examples.
Nature holds my camera: they tried visitor blogging with a terminal in gallery so people could ask questions.
They talked about the IMA dashboard. [I asked a vague question about whether there was a user-driven or organisational business case for it – turns out it was driven by their CEO's interest in transparency, e.g. in sharing how they invest monies, track stats and communicate with their visitors. It helps engender trust and loyalty e.g. for donors. Attendance drives corporate sponsorship so there was a business case. It's also good for tracking their performance against actual actions vs stated goals.]
The advantages of using a web application toolkit – theromansarecoming.com took $50,000 to build for a four month exhibition. It hit the goals but was expensive. [The demo looked really cool, it's a shame you don't seem to be able to access it online.]
Breaking the Mode was built using existing components on the technical side, but required the same content investment i.e. in-house resources as The Romans Are Coming. The communication issues were much better because it was built in-house – less of a requirement to explain to external developers, which had some effect on the cost [but the biggest saving was i re-usable component] – the site took 25 hours to build and IT staff costs were about $1000. [So, quite a saving there.]
They demonstrated 'athena', the IMA's intranet. It has file sharing and task management and is built on drupal, looks a bit like basecamp-lite but without licensing issues. "Everything you do in a museum is project-based" and their intranet is built to support that.
There was discussion about whether their intranet could be shared with other museums. Rob Stein is a firm believer in open source and thinks it's the best way to go for museum sector. They're willing to share the source code but don't have the facilities to support it. There's a possibility that they could partner with other institutions to combine to pay small vendors to support it.
[I could hear a sudden burst of keyboards clicking around me as the discussion went onto pooling resources to create and support open source applications for stuff museums need to do. Smaller museums (i.e. most of us, and most are much smaller than MoL) don't have the resources for bespoke software or support but if we all combined, we'd be a bigger market. Overall, it was a really good, grounded discussion about the realities and possibilities of open source development.]
Back to the slides…
Team Troubles
[It was absolutely brilliant to see a discussion of teamwork and collaboration issues in a technical session.]
Divide and conquer – allow team members to focus on area of expertise. Makes it easier to swap out content and themes.
They're using MVC – Model (data management), Controller (interaction logic), View (user internface). They had some good stuff on MVC and the web in their slides (around 77-79). They also discussed the role of non-technical team members.
Drupal boot camp
[This was a pretty convincing demo of getting started with Drupal and using the Content Construction Kit (CCK) to create custom content types e.g. work of art to publish content quickly, though I did wonder about how it integrated with ORMs that would automatically pick up an underlying data structure. Slide 103 showed recommended Drupal modules. It's definitely worth checking out if you're looking for a CMS. If you're on Windows, check out bitnami for installation.]
Client side development
"The customer is always right"
They talked about the DOM (document object model) and javascript for Web 2.0 coolness.
They recommended using Javascript toolkits – more object-orientated, solve cross-browser issues, rapid development. Slide 109 listed some Javascript toolkits and they also recommended Firebug.
Interface components
They should be re-usable, just like the server-side stuff. They should some suggestions like reCAPTCHA, image carousels and rating modules. Pick the tools with best community support and cross-platform support.
CSS boilerplates
Treat CSS like another software component of web design and standardise your CSS usage. Use structured naming for classes and divs in server-side content generation. Check out oswd.org for free templates.
XML in the real world
They demonstrated Global Origins (more on that and other goodness at www.ima-digital.org/special-projects) which uses XML driven content.
Questions and discussion
I asked about integration with legacy/existing systems. Their middleware component 'Mercury' binds their commercial packages and other applications together. e.g. collection management system extraction layer. [This could be a good formalised model for MoL, as we have to pull from a few different places and push out to lots more and it's all a bit ad hoc at the moment. I think we'll be having lots of good discussions about this very soon.]
Some discussion about putting pressure on vendors to open data models. It's a better economic model for them and for museums.
Their CEO is supportive of iteration (in the development process). The web team is cross-department, and they have new media content creators.
[I was curious about how iterative development and the possibility of making mistakes work with their brand but didn't want to ask too many questions]
They made the point that you have a bigger recruiting pool with open source software. [Recruiting geeks into museums has been a bit of a conference meme.]
They give away iPods for online surveys and get more responses that way, but you do have to be aware that people might only give polite answers to survey questions so pay close attention to any criticism.
The IMA say you should be able to justify the longevity of projects when experimenting. Measure your projects against your mission, and how they can implement your mission statement.
So, that's it! I hope I didn't misrepresent anything they said.
Great post, so thanks heaps. I've passed this to our web geeks who will understand much more of it than i do!
Mal