I posted all data I managed to collect on WPF/e vs DHTML vs Flash vs WPF performance test results to Swivel for analysis and visualization. Selected charts are embedded below — enjoy! Thanks to everyone, who posted his results — it made up total of 73 measurements, which I could never do alone. Along with fps itself I used fps / GHz parameter to be able aggregating results from different machines. I realize that Core 2 Duo GHz was not born equal to AMD Duron GHz but it’s better than nothing.

I also realize that making statistics from such a small dataset could be misleading. However aggregated results seems to agree in general with what I observed myself so I guess there is nothing wrong with them.

Continue reading…

A while ago I posted an article titled Walls and Towers discussing the “language divide” problem of the internet.

It was amazing to learn that there are projects out there, taking seriously on the problem. One of them is 10 years old (!) project of Brian McConnell called Worldwide Lexicon. Its latest incarnation is targeting to provide bloggers and websites with combined automatic / human translation services. The project is open source and public beta is expected soon. It sounds great and I can’t wait to see what will come out.

Another project Kontrib is a recently launched clone of Digg (sic!) with integrated machine translation. It “speaks” English, عربية , Español and Français for now from which I only know English so can’t really test it but some Arabic to English translations look intelligible enough (compared to the original). The site is kinda funny looking too but it’s forgivable if it works.

Yes, it is I learned, when upgraded my laptop’s hdd from 60Gb @ 7200 Hitachi drive to 100Gb @ 5400 Toshiba one. Despite the fact that Toshiba is new and shiny model with 16 Mb buffer, it’s considerably slower than three years old Hitachi (ex IBM). Toshiba costs $75 now and Hitachi was probably $200+ at the time it was made.

Having said that I don’t notice much difference without instrumental measurements (I used HDTune utility for them).

The measurements show that where I had 14.3 ms access time with 7.2k, I got 17.4 ms with 5.4k. (It’s 3 ms difference, which may seem a little, but it’s also 20%, which is a lot and matches difference in rotation speed between devices precisely). Similarly transfer rate of Toshiba is 27.2 MB/s (max 34.3 MB/s) while it was 29.7 MB/s (max 38.5 MB/s) for Hitachi.

Upgrade was simple as I used a trial version of Acronis Migrate Easy utility to make exact copy of my old partition on the new drive. It worked great and moved 50 Gb of my data to 73Gb partition on new hdd in less than 40 minutes. The new drive booted just like nothing changed. I reserved 20Gb for “experimental use” and will be running Linux(es) on them.

This week I was surprised to learn that nVidia support download of nForce4 Windows Server 2003 x64 drivers includes no 64-bits drivers at all!

You should use Windows XP x64 drivers instead and even with them I didn’t manage to install the system on SATA RAID. Windows setup accepted drivers from the ancient floppy disk (ridiculous procedure by itself in 2007) but ended with blue screen of 0×0000007B Inaccessible Boot Device error.

I used Adobe’s Apollo Alpha 1, which was released today, to package Flex and DHTML versions of my web animation benchmark. The goal was to try the thing in the action and see how its performance compares to Flex and DHTML animation running in the browser.

Apollo is really cool (see screenshot below) and packaging both Flex and DHTML applications with it is very easy.

Here are some performance results (compared with the same code running inside the browser), which I got on Pentium M 1.7 running Windows XP SP2.

Animation performance in frames per second (fps); higher fps means faster performance.

* Unstable

Please note that Flex version of the test was running with transparent window. HTML-based Apollo applications doesn’t support this (just like any other web-browser).

Although Apple WebKit in Apollo scored 98 fps it looked like its real refresh rate is less than 20 fps and it simply skips rendering of most frames. Opera, while showing the same number, feels much smoother.

I also checked memory consumption and while both applications start with ~30Mb RAM usage the growth rate (yes, memory leaks, which is of course fine for alpha) is quite different. Flex-based Apollo app does only ~1.2 Mb/minute while HTML-based makes as much as 10 Mb/minute.

If you have Apollo runtime installed you can try both applications:

Flex-based bubblemark (download source code)
HTML-based bubblemark (download source code)

Yes, it’s true and here is the link http://www.crackcity.ru/.

Check it out and you’ll like what you see. The site is built by PR-agency FMC Group and intended to convince users buying legal Adobe software. The idea is interesting and the site looks authentic although there are couple of problems with it:

• This won’t work because you can’t get blood from a stone. I can’t imagine a user, who is impressed with this trick to the degree that will make him go shopping for Adobe Creative Suite instead of looking for cracks further
• The site actually helps finding cracks for Adobe software. To make it more authentic FMC people used real filenames of crack releases. Everybody who ever tried searching files with Google knows that finding the file by its exact name is much easier than using generic queries
• The anti-piracy site violates 3rd party copyright itself. The quadratic easing in equations code used in JavaScript on the page is copied in violation of the license from Robert Penner’s ActionScript code.

  Continue reading…

There have been quite a few discussions lately, which deal with SaaS economy and mostly everyone agrees that to fully leverage SaaS advantages we need a platform. The question is who will be that platform. The runners include both “old school” like Salesforce.com and Netsuite and newcomers like Apprenda plus uncertain (but surely non-zero) number of companies “under the radar”. The point I want to make is that pure platform players have much less chances to win than those who came to the platform space from offering products of their own.

To make the platform succeed you have to build your product on it first. Even turning your product into the platform is better than just making the platform. Why so? Why can’t you just play pure platform game and be just enabler without making your hands dirty in offering actual products? Isn’t pure way better thanks to the absence of interests’ conflicts inevitable when the company sells both the platform and products? There are numerous reasons; I’ll try to address some of them below.

Continue reading…