OSCon day 2 : Tutorials

• Learning AJAX - Alex Russell
  • Most interesting stuff I learnt Size of the data you can send and cache in a typical XMLHTTPRequest : 4K (small!). Best data type to return : JavaScript instead of XML.
  • Take away Good but underwhelming presentation. The talk was mostly on good old DHTML (JavaScript and DOM), few AJAX specifics (best practices in XMLHTTPRequest use). Because of browsers limitations (browser bugs, browser data set issues, tools), it doesn't look like AJAX is ready to be used to create high performing destop like applications. Note though that, if judged by the attendance, AJAX is really hot : the room was packed full!
  • Details
    • Why WebApps are cool? No deployment headache, zero install...
    • "Web App is a limiting paradigm", time to think about those things as "applications"
    • Cache size (for XMHLHTTPRequest return packet) : 4K (think about it when building your app...)
    • XMLHTTPRequest calling sequence changed over time. Some hacks required to make it work in all browsers (not too big of an issue though if you limit yourselves to most recent browsers...)
    • Tip : return JavaScript instead of XML (or HTML) : less data required to hold for instance large data sets (less markup cruff), parsing easy on the client (eval() is your friend).
    • JSON (data packing format) : ? something to check...
    • Tool : get Rhino from Mozilla (JavaScript standalone running environment and command line)
    • Tool : use a protocol analyzer like Ethereal
    • Tools in general (or the absence thereof) are an issue with AJAX ('cause lots of bits and pieces of standards involved).


• Introduction to XSLT - Evan Lenz
  • Most interesting stuff I learnt XSLT does work in most modern browsers. Use it instead of killing yourself with weird JavaScript to add markup to XML data.
  • Take away Great talk, very clear, solid presentation by Evan though somewhat unwiddling in the end. Lots of participation from the floor. I did finally understood something of XSLT! Seems like a great subtitute to JavaScript code to create markup around XML snippets. I wouldn't use this to style though (as shown by Evan), let's use CSS for that...
  • Details
    • XPath part of XSLT (though XPath used also in other places like DOM and pretty much in any W3C spec where there's a document tree).
    • Use of XSLT : XML to HTML document generation, *ML to *ML conversions (e.g. OOo XML to Chandler parcel XML could be interesting to us?)
    • Can do client side XSLT in most browser
    • Python XSLT engine : see 4xslt
    • IMHO, using XSLT to style XML docs does not seem like a good idea (leave that to CSS)
    • Use template rule preferably to flow control rules (also possible in XSLT)
    • Text nodes are tricky! Space and line feed do count as text nodes... Use? Round tripping and transform without blowing up the formating...
    • Can use predicate on XPath to filter lists
    • Core processing model : template rules
    • XPath provides lists (or nodes sets) from trees, XSLT processes lists and spits out trees
    • Beware of priorities when selection conflicts (only one rule can apply per match). You can set priority if needed to overrule the default one.
    • Beware of XSLT built in templates rules! They turn everything to text, not identity.
    • Learn to write and use the XML to XML identity transformation
    • Tool : Xselerator 2.6 : XSLT IDE and debugger from MarrowSoft