JCooney.NET

Leon forwarded me this gushing NY Times regurgitation of this apple press release. This part on upcoming advancements in OSX Snow Leopard’s use of mutiple cores was interesting:

Snow Leopard delivers unrivaled support for multi-core processors with a new technology code-named “Grand Central,” making it easy for developers to create programs that take full advantage of the power of multi-core Macs.

The NY Times quoted Steve Jobs as saying:

“The way the processor industry is going is to add more and more cores, but nobody knows how to program those things,” he said. “I mean, two, yeah; four, not really; eight, forget it.”

And Apple’s “snow leopard” landing page says this:

Grand Central takes full advantage by making all of Mac OS X multicore aware and optimizing it for allocating tasks across multiple cores and processors. Grand Central also makes it much easier for developers to create programs that squeeze every last drop of power from multicore systems.

Colour me sceptical. Concurrency is a hard problem and current programming models rely on the developer to (try and) keep track of what is shared and what isn’t. Apple is certainly not the first company to notice the rise of multi-core, and not the first company to try and ease the pain for simple-minded developers. On top of the “normal” concurrency stuff (threads, monitors, locks etc) that are built into .NET Joe Duffy and co have been cranking out parallelfx releases for at least 6 months (and discussed for nearly a year and a half). Joes upcoming book on concurrency looks awesome. I’ve done some sholder-surfing while Joel has been flicking thru some of the chapters in the rough cuts version and I can’t wait. You can pre-order form Amazon.com here.

One thing I did like about the Apple press-release was the announcement that they are working on “foundational” enhancements in snow leopard rather than a lot of user-facing features. This is something that Microsoft tried with Vista. Time will tell if it has paid off for them or not, but I certainly think there is a lot of work that could be done on “foundational” stuff and “fit and finish” in Windows, rather than going for bullet-points-on-the-back-of-the-product-box features.

Comments

Paul recently wrote a great blog entry that called out something I’ve felt for a long time, regarding the absence* of an editable datagrid in the “out of the box” WPF control set. While many hard-core winforms users bemoan the lack of a data grid in WPF, I think the absence of a data grid in WPF is a good thing. Why? To me the editable data grid is an over-used UI device, which shows either a terrible lack of imagination or understanding as to how data is going to be entered, or an abdication of UI design responsibility. To be clear - I’m not talking about displaying of tabular data ala the ListView (WPF’s list view is very powerful for this). I’m talking about banging in data, row by row, cell by cell. We don’t give users a copy of SQL Management Studio, show them how to open up underlying tables and let them go nuts, but (to me) giving them an editable grid is not that different.

The reason editable data grids fall short is that when a user enters some data that means something, and I believe data grids don’t do a great job of capturing that meaning because all data entry operations look the same. A bank employee changes a customer credit rating from “good” to “credit risk” using a drop-down list in an editable grid means the nature of that customer’s relationship with the bank is about to change fairly dramatically. Why was the change made? What validations need to be applied? What other processes need to be initiated as a result of this change? As Chris Stevenson pointed out a while ago, editable grids don’t do a great job of capturing these subtleties. Data grids are the “one size fits all“ data entry solution you don’t even have to think about. Sure, the data you get may not be as good as it could have been. The user might have had to jump through a few hoops to get what they wanted done, and they may have been exposed to some implementation details of the system but the data got in there.

Users and analysts will often request an Excel-like interface to data. Some people think that given Excel’s ubiquity this will reduce training. However Excel is a commercial product with hundreds of person-years poured into its code-base. Is the grid you’re using going to be as fast as Excel (given that unlike Excel your data probably has to come from a remote server)? How will your grid handle concurrency (Excel doesn’t have to care about concurrency)? Excel has hundreds of features and tweaks to make editing of tabular data faster - how many of those are you going to implement? Which ones? When Excel v.Next comes out will you rev your application to continue to “be like excel“? If end users want Excel why not just let them use Excel?

When the only tool you’ve got is a hammer, everything looks like a miniature in-line version of Excel. By keeping the data grid off the table for a while the WPF team may have forced people to get a little more creative in how they interract with their data. At the end of the day “it that makes me find creative ways to work around its deficienceis“ is not a great selling point for a framework (although it never seemed to hurt the adoption of HTML) and the WPF team have come to the party and said they will provide a grid. That still doesn’t change the fact that I think they are generally a bad idea.

* There are 3rd party WPF grids from Xceed and Infragistics (both of which are free IIRR).

Comments

»

seen here »

Super-smart-guy Peli disappeared into MSR a while ago to work on Pex, an exploratory testing framework for .NET. There had been lots of teaser blog entries, a few conference appearances, chats with Hanselman, a video or two but no release, until last monday when Pex 0.5 was released. Testing Supremo Bruce McLeod is planning a series of posts explaining all of this tentatively titled “the Joy of Pex”. He then plans a follow-up article “how to explain to your spouse why you’re staying up a night reading something on the internet called ‘The Joy of Pex’”. »

I just finished delivering my WPF talk at ReMix melbourne, which seemed to go OK. The talk was on WPF 3.5 and 3.5 SP1, and was somewhat based on the talk that Rob Relyea delivered at Mix08 in Las Vegas. Here’s the slides, and the demos (one of which was borrowed from Kevin Moore here, and the pixelshader one is based on Greg’s demo here). Leave me a comment if you have any feedback.

Comments

I had a look at the new World Wide Telescope created by Microsoft Research today. Like scoble I also wanted to cry when I saw it, but perhaps not for the same reason. Unlike scoble (who did it for publicity) I felt like crying because it seemed like what COULD have been a great showcase for WPF instead stooped to “legacy” technologies. Please MS - WPF was made for creating these kind of apps. If you’re not going to  use it to build things like this then why should anyone else?

Comments

Here are my slides from the talk I gave at CodeCampOz 2008 on “Silverlight 2.0 and WPF - what’s the same and what’s different”. Thanks to everyone who came along to my talk. I’ll be doing the same talk in June at the QMSDNUG in Brisbane. I had to cancel on Mike for the May session because of ReMix, where I’ll be talking about WPF 3.5 and the coming improvements to WPF.

slides as PDF 691K

Comments

I received a gentle reminder this evening that interfaces aren’t contracts. I was attempting to bend Silverlight 2 to my will, specifically the ResourceDictionary. I really wanted the keys, and was pleased to see that it implemented IDictionary<object,object>, which would give me a collection of keys (of object). I was dismayed to see the following implementation in reflector for the Keys and Values properties.

    ICollection<object> IDictionary<object, object>.Keys
    {
        get { throw new NotImplementedException(); }
    }

    ICollection<object> IDictionary<object, object>.Values
    {
        get { throw new NotImplementedException(); }
    }

This really brought home to me that Interfaces are nice and all, but really only tell us that a type will have methods with a certain name that take certain paramters - exactly the same thing the compiler tells us about lots of other things (that is if you’re in a compiled world….the benefits of interfaces seem even more marginal in more dynamically typed duck typing* languages). It doesn’t guarantee that the methods will actually DO what we think they’ll do (or even that they’ll do anything at all except throw an exception in your face). Don’t get me wrong - I like and user interfaces, but they’re not a contract, more of a gentleman’s agreement or shared understanding. Now if you’ll excuse me I have some nasty reflection hacks I need to brutally force on the Silverlight run-time.

* Aside: speaking of duck typing, I heard of another interesting turn of phrase for monkey patching - duck punching! If it walks like a duck and flaps like a duck but doesn’t quite quack the way you want it to you just punch it [chewie].

Comments

I was reading through a discussion thread on a programming forum, where someone was proposing that a large software vendor should pay community members for ideas on how to enhance said software vendor’s platform. This is an interesting idea, which files in the face of my own personal belief that ideas (with a few notable exceptions) are worth approximately nothing. It is the implementation/execution of the idea that is worth something. Lets look at some big software companies - Microsoft and Google. Operating systems were certainly not a new idea when Microsoft released MS-DOS, it was their implementation of the idea and the royalty stream they created through savvy licensing to IBM that made MS. The operating system they licensed to IBM wasn’t even theirs, it was a CP/M clone they bought for $50K. Internet search and on-line advertising were not new ideas when google entered this market, but their dogged refinement of their search implementation built the brand/trust in the company, which in turn lead to extreme user buy-in and eventually a strong “network effect” between on-line content providers (AdSense) and advertisers (AdWords). Strong implementations of ideas are bought and sold every day (they’re called companies, or the rights to products) whereas there is not currently an established market for ideas.

See also: No One Is Going To Steal Your Brilliant Idea

Comments