In the afternoon session today, I’m going to learn what it’ll take to build the next generation of apps. I’m curious to see what they’ve got to say.
Presenter: Dion Hinchcliffe
- Room is about 90% full, compared to previous session
- Poll shows 45% tech architect/dev, 45% product management, 10% other
I fear that this might be more beginner than I thought. This seems to be covering a lot of existing Web 2.0. I hope it’s just to set common background.
- Need to look into social graphing as a metrics method
- Badges are becoming most popular form of widgets (e.g. Flickr, but I have no idea what this is, need to find an example)
- W3C trying to define the Web Widget as a standard, likely to be undermined by the industry who will not wait for them to get it all done
Yep. Basic information, mostly. I gather a lot of people here are fairly new to all this. Not as bad as the introductory SES stuff we had, but still…
- Some service providers have rate limits (e.g. Google Maps) after which you have to pay; Amazon requires it from the get-go
- Basing yourself on other services has drawbacks: - Weakest link syndrome (e.g. Amazon S3 outage)
- Limited economy of scale due to dependency on others
- Requires as good legal skills as dev skills to handle SLA issues
- Top APIs available on ProgrammableWeb.com (which is partnered with IBM)
- Twitter built on Amazon tools
Okay, minor rant: Woman just in front of me asked if I “had to do all that typing”. For crying out loud, lady, this is a Web 2.0 technology conference, not a press conference for Fergie. SUCK IT UP!! I can’t help it if I learned to type on a classic clunky IBM keyboard and have had hard keystrokes ever since.
- Need to look at Yahoo! Geocoding as an alternative for Google’s
- Platform solutions tend to beat an application for ROI and customer support every time
- Amazon EC2 allows rapid roll out of virtualised servers - Second level Amazon apps (based on Amazon services): Jungle Disk, Digital Chalk
- There is no proven ROI on social networking apps (e.g. Facebook app)
- Numbers in XML data harder to process, increases parsing time — use JSON for numeric data instead
- Dion believes REST is better than SOAP (mostly due to ease in parsing REST data)
- Google Gadgets - 47k widgets available, all through iGoogle (and outside use?)
- OpenSocial based on Google Gadgets
- WidgetBox - 50k+ widgets
- 2.5m serves of the widgets
Lots of discussion on mashups. A lot of “we know that already” info here. One interesting piece: Facebook and Amazon have an interesting partnership that allows someone to create a Facebook app, host it through Amazon, and rollout a grid-served application with no infrastructure for little cost.
- Major APIs (in order) - RSS
- REST
- JSON
- SOAP
- ATOM is upcoming but not in the top 4 (ATOM is REST-compliant, so might overtake SOAP before long)
Long section on AJAX. Sorry, I mean Ajax. Nothing we haven’t done at least a dozen times here.
- Dion a big proponent of Silverlight, but seems to be missing some details - Microsoft rep told us Silverlight could not yet handle 3D outside of Windows due to rendering engine; similar note about HD video
- Either doesn’t know a lot about Flash, or is unwilling to draw comparisons for some reason — much of what he claims as awesome about Silverlight is already in Flash
- Agree that Flash needs to go Open Source to eliminate the last part of the RIA puzzle (only thing that’s still wholly proprietary)
- JavaFX is Sun’s RIA tool (never heard of it until now); seems to be geared mostly to mobile - Scripted, not compiled
- Does not replace Swing, but makes it easier to develop
- Can be used to create arbitrarily large apps
A note on RIA and metrics: Using asynchronous apps kill the pageviews metric. (Yahoo! got burned for this a while back.) Need to make sure that we account for this by tagging sub-apps and/or data.
New developments - New tools are great, but you lose performance with greater abstraction (e.g. Ruby)
Ruby on Rails is great, runs Twitter; massive performance drawback
- IBM has proved 10-20x savings in initial dev costs over traditional tech, like Java
CakePHP is a framework written in PHP, modelled after Ruby on Rails (but not a port)
Groovy & Grails is a Ruby on Rails-like framework for Java
- Does not have performance drawbacks of Ruby on Rails
Could be a hot one to keep an eye on, and run tests with
Platform As A Service (PaaS)
- Google App Engine
Amazon AWS, EC2 (Elastic Server on Demand)
Heroku
Keep sight of the goals - Watch your competition (don’t replicate what they did, you need to outdo them)
Know your customers
Use the right tools for the right reasons
“This is an arms race”