Dan's Developer Diary (3D)

danielruspini1 at 10:29:23 AM PST Permalink | Blog about this entry |  Add to del.icio.us |  digg this This entry has 0 comments: Add your own

 

Dojo 0.9: Faster, Leaner and Dijit?  (Dylan Schiemann - Dojo co-creator)

This session was cool because although I'd heard a lot about what was happening with 0.9 this was the first time I'd seen some of the changes and real size/performance numbers.  The idea behind 0.9 was to scale down the code which had grown bloated over time due to the nature of the core principles: support as many browsers as possible, allow contributers to add widgets to the widget library, support backwards-compatibility code.  What they found was that 30% of the code was expendable.  Also the widgets had been written in a variety styles and were not all consistent in use; while others were not used by most developers.  In some cases very popular widgets were trying to do too much by covering every possible use case which made them slow (ex. Button). 

 

Dojo 0.9 is a rearchitecture in that many of the namespaces have changed, the API trimmed (ex. won't support 14 different args to connect an event), and there will be no backwards-compatibility support for 0.4x.  There will also be no v0.5-v0.8, the idea is for 0.9 to be the beta for 1.0 which should be ready before the end of 2007.  They've also had the chance to integrate new ideas and rework 0.4 pain points (blank page problem, module overload, "spaghetti requires", markup parsing performance, widget layout performance, spotty theme support, inconsistent documentation, too much "magic"/cleverness).

 

Dojo 0.9 is now 3 top-level projects.

• Core - The main API (Dojo Base)
• Dijit - Most used widgets (all support I18N, accessibility, theming).  Also notable is removing templateCSS support, now all CSS will be together which should make for less CSS.  New Tundra theme across widgets, they all look the same out of the box!
• DojoX - Experimental or less used API. (dojox.gfx, offline, comet, charting, event advice)

They've dumped Ant in favor of a Java-based process which works well with Rhino.  I think we've already done a lot of this for Webmail, but you can build your code, not just Dojo code, can "intern" templates, create layered builds.

 

Performance!

I’ve copied this all from the slides, so feel free to look at them in pretty graph form here:

http://ajaxexperience.techtarget.com/images/Presentations/Schiemann_Dylan_Dojo_0.9.pdf

 

Build Sizes

Variant	0.4 “Ajax”	0.9 Base
Uncompressed	319K	169K
Built	148K	43K
gzipped (OTW)	39K	20K

 

More Build Sizes

Variant	0.4 “Widget”	0.9 Base + Digit
Uncompressed	443K	233K
Built	197K	61K
gzipped (OTW)	53K	29K

 

Other interesting numbers

0.4.3 vs. 0.9 Object Allocation

“Ajax” build = ~30% less

Widget infrastructure = ~30% less

100 widgets = ~60% less

 

0.4 vs. 0.9: Micro Benchmark

100 buttons (programmatic) and 100 buttons (Markup)

Look at the PDF for somewhat exact numbers, but suffice it to say they look VERY good…drastic improvements

 

We’re now in the under 1 second category when it used to be upwards of 10 seconds.

 

Results

50-66% Smaller

5-20x Faster

35-50% Fewer Objects = less memory and CPU utilization

 

 

Furthermore, there are a few new parts of Dojo that will help us rely less on widgets.

 

Introducing dojo.behavior

Not everything needs a widget

Progressive enhancement

Apply JavaScript to CSS selectors

 

dojo.query

The basis for widget parsing and for behavior

Accepts most CSS3 selectors

Supports scoped queries

Returns a “real” array

Really, really fast

Comparable to jQuery

 

Finally what good would a new dojo be without a new powerful new testing tool….

 

Testing: “D.O.H!” The Dojo Objective Harness

D.O.H (think Homer Simpson) is the new unit testing platform.  Can be used command-line or within the browser.

 

My final thoughts

I ran into Alex Russell (co-creator Dojo) in the hall and asked his opinion on how much time it would take to port a dojo RIA to 0.9, and he told me around 2 days.  So translate that to a bigger app like Webmail and I'm hoping that we're talking realistically like 2 weeks.

 

Feel free to send me questions and if I can’t answer them I’ll make sure they do get answered by some one inside the Dojo community.

danielruspini1 at 3:37:45 PM PDT Permalink | Blog about this entry |  Add to del.icio.us |  digg this This entry has 0 comments: Add your own

Walking away from the Grand Hyatt last Friday I had the feeling that although the developer community is large, there is a strong sense of unity and direction amongst us.   Throughout the 3 day conference the defining theme was that we are all in this together and we have the opportunity to pave the road in front of us.
On the last day of the conference I attended the Future of Ajax panel discussion.  On stage were Brenden Eich (Mozilla), Douglas Crockford (Yahoo!), John Resig, Bill Scott (Yahoo!), and the Ajaxian MCs Dion Almaer and Ben Galbraith.  They participated in prepared discussion topics as well as answer questions from the audience.  What struck me more than ever before is that developers and the people who create the browsers were talking about where to go next, what works, what is broken, and who was doing what to make progress.  A comment was made that developers have pushed the boundaries of the browser once again and that the browser makers need to catch up.  Between the remarks of Chris Wilson (Platform Architect of Internet Explorer) and Brenden Eich I was certain they were listening to the audience in regards to what standards should be and where they might concentrate their efforts.  For the first time ever it seemed like we were all in the same game, the browser companies were not these monolithic faceless entities shipping browsers that developers had to create hacks for to make work the way they wanted.  Finally, although still perhaps a few years away, we've agreed to work together and get over the hurdles so we can enter the next phase of web development.
(jQuery/Mozilla),

danielruspini1 at 11:34:36 AM PDT Permalink | Blog about this entry |  Add to del.icio.us |  digg this This entry has 0 comments: Add your own