@Sven Groot:

Yeah it does look like error on my part as I don't have a comment there explaining why I have made it more complicated that it should be. I'll get back to this if it starts to fail at runtime later.

I use a serialization lib which wants this:

public static void Serialize<T>(Stream destination, T instance);

and I want to call Serialize(stream, ????);

Such that the ???? part is not manually typed in but inferred from the type of the list variable.

How to do this?

var list = new List<KeyValuePair<string, decimal>>();
myClass.GenericMethod<typeof(list)>(list);

I'm pretty sure C++11 had something like this, why doesn't C# have it?

http://www.moroccoboard.com/viewpoint/59-adil-naji/1173-moroccan-banks-get-fat-on-expats-money-transfers

I came across this quite bit accidentally, it's a pattern that's been going on for years. Didn't someone claim the markets are a random walk? I think this proves that to be incorrect.

I haven't really investigated but I doubt whether you can get the quoted rates - as they seem to be a key part of the scam explained in the link, used by the bankster mafia to skim money, in addition to various fees they take in addition of the bid/ask spread (according to the article they skim/rob you atleast 4 different ways). The trick is that when you exchange money they pick a date from now to few days back that gives you the worst possible exchange rate or something along those lines according to the link.

Is Google allowing 3rd party bitmaps in their advertisements but hosting them on the gmail server?

I could speculate that it was perhaps some 0day attack from a 3rd party bitmap data hosted by gmail, but I opt not to speculate such things, so lets leave that theory at that.

http://judy.sourceforge.net/application/shop_interm.pdf

Here's a javascript compressed trie benchmark.

http://lookups.pageforest.com/test/perf-test.html

I'm using IE10's old preview and for some reason that preview is abysmally slow, must have been fixed since as the other IE10 results aren't that bad. 

I'd be very interested in speculation of whether you think AVX2 with its new bit manipulation (bit scatter/gather, shifts etc) instructions would speed up operations related compressed suffix arrays / tries. I skipped over the Ivy Bridge upgrade and preliminary Haswell tests @ Toms suggest the most interesting speedup would be from AVX2 optimizable code.

Very simple logic: If x time (15-60 min) has passed since the first "submit" of a new thread/reply, then make the it public. Also don't publish the edits until after 1 minutes, so I can go and edit the edit (typo fix for example). So If I kept editing the message and submitting edits every 59 seconds, it would never publish the thread/reply until I stopped doing that.

I found that there is currently a problem with this (it was working decently previously, but I can't recall when I last edited intensely here).

symptoms: I posted a new thread, then tried to edit and the edit box shows message contents as either :

1) as html code (without pressing edit as html)

2) or if I did several edits in one minute, it displayed some old contents (second last edit or "previous to previous"), so if I had saved again there would have been data loss.

The both problems occurred at the time I tried to edit:

http://channel9.msdn.com/Forums/Coffeehouse/C-Sux

So look at the time stamps of when I clicked edit on that the first 2-3 times today.20130321 04:20 EST approx.

http://sux.di.unimi.it/

https://github.com/simongog/sdsl

The important factors in these is of course the amount of memory taken and the time to construct/search the compressed array/index.

It may well be that it mkaes most sense to provide managed interfaces to C/C++ lib that construct the tree and perform the traversal but I don't know for sure yet how much slower C# is these actions.

Encodes system audio on fly to mp3||aac||ogg with slow gating, interleaved with variable frame rate video where frames are generated based on changed pixels on the screen and the delta is stored in such a way that when playing and seeking in the file, the seeks are made to the audio stamps and then the image is rendered by backtracking to "keyframe" and forwarding through the deltas to the audio stamp.

So if you capture say infinite hours of audio and video where nothing happens and ~5 seconds of audio buffer statistically appears to be just noise, the resulting playable video file size would not change until something "statistically significant" occured on the audio or video.

Something like saving the stream from a "remote desktop" protocol to disk might do the trick as long as it could be implemented locally without enabling remote desktop.

Requirements for blocking:

Software which does not have administrative privs should only be able to capture its own window content and if it's a browser then the any activex plugin should be restricted by the browser to be able to only capture the tab the plugin loaded from while the user visible URL stays the same. I don't really care if existing apps break due to this change, those apps that break could be whitelisted in HKLM.

(In these comments I just looked at with the IE10 preview, there vertical lines are now 1 pixel column black and another red, so it just looks bad)

By using a good quality CRT (reads fine->) in 1600x1200, and then in windows 7 setting "make text and other items smaller/larges" control panels "Set custom size (DPI)" to 130% and selecting "Use Windows XP style DPI scaling", in the IE10 preview that I have installed text actually becomes much like GDI text. (Zoomed in 300% to look at the pixels for a font known to be unreadable in IE9/10 normally).

Of course this means I don't quite get the full use of 1600x1200 and it looks more like something near 1280x1024 desktop, but atleast the text is quite readable and the only thing I'd like is a CRT with 3200x2400 resolution or more at maybe 25" size.

The key here is that the DPI zoom % should be set so that the most commonly seen worst scenarios with IE9/10 font rendering actually become non-issue. Where normally at the default setting the "i" "l" "m" "h" etc have the vertical likes rendered with only blue and red color (seen easily zoomed in mspaint and just look fuzzy when not zoomed in), at around 130% zoom they start to be rendered mostly normally. I didn't test many cases though yet , just the couple worst cases I've seen.

1. User can create a C++ native project and add C# classes directly in the same project next to C++ classes. The C++ code can be compiled with 3rd party compiler like Intel compiler and magic happens which allows the C# classes to talk with the C++ classes (and invoke C functions) and the other way around as if they were the same language, no performance penalties at all.

2. The C# classes code in the C++ project can be modified while the entire application is running in optimized release build. This happens by attaching to the running project, suspending it for a millisecond when it hits a pre-determined "safe to hotpatch section" and updating the C# code inside.

3. Changes to C# portions do not cause the C++ code to re-compile or link.

4. As alternative to embedding C# the embedding interface supports embedding other future languages which gain the same features as above.

Since I'm not a compiler developer I don't see any technical problems here.

Maybe the keyboard piece could contain the main battery and you could detach it and plug in from the other side, such that the keyboard+main.bat. pieces keys would face the back of the part containing the display, and the keyboards underside would be the bottom of the device when used as a tablet. To facilitate this transition, there would be a smaller battery inside the display containing piece than supplies power while you are switching the keyboard+mainbattery to the other side (laptop->tablet form). I haven't really deeply thought about this though so there could be some issues I haven't anticipated. I just noted eg. Anandtech complaining about the rigidity when used legs crossed, if the keyboard contained the main battery this would distribute the weight better and provided more rigidity.

I would like to have an option to instead view the array in similar view as the "memory view" with 16 hex columns and 16 ascii or unicode side by side.

This should come from hover + right click on the myBuffer and selecting the option.

Also I'd like the following, just as a shortcut to view the memory, or is there a hot key to do this when you are hovering over a variable?

Middle click of the myBuffer, should open the memory view or similar, range bounded (not allowing to scroll to earlier than or after the myBuffer memory) to the myBuffer memory and with option to unbound the range, so memory prior and after the buffer can be scrolled into.

(Again, no idea why it's 10-15 FPS in IE8 while viewed embedded here)

I like that the author chose an approach of "reviving" the disabled start menu code existing in Windows 8.

However even with these fixes in place, Windows 8 still comes with non-trivial under the hood changes which in my interpretation risk compatibility and performance changes of old applications. It's most prudent for me to wait for those who have time time and interest to sort out "bleeding edge issues". Windows 7 didn't come with such major changes and in Vista timeframe Channel 9 had done great job in communicating things ahead of time so I came in prepared to Vista, the only "show stopper" for Vista in my case was that Explorer bug that AFAIK still has not been fixed and made Vista completely unusable as a serious OS. Since that bug is gone with Windows 7, that just proves that it's a Vista problem and MS didn't care enough about Vista customers to fix a show stopper in SP. That's where Microsoft lost a great deal of trust, next step was IE9 fuzzy fonts. No fixes there either despite that being widely reported in good time before release. And then of course came all the crap with VS 2010 performance and 2012 looks and now the Windows 8 mess.

Automated testing may be wonderful but it won't find perceived, visual or usability problems or things that the test automation programmers couldn't think of. If one doesn't take care to listen to user feedback about prior to release and address those one way or another, either the complaining continues or it will stop, which means that the complainers stopped caring about your product and moved to another product or won't upgrade anymore. I have been on the "don't care/upgrade" but since MS is now at work on next release after 8 is RTM, complaining may resume, as the alternative is to jump ship and I don't care about switching one set of OS issues to another + having to learn "new layout of piano keys" so to speak.

Restored Desktop to its place With StartIsBack, computer starts to desktop. Always. Unlike other tools, which 'skip' Metro by sending keypresses, StartIsBack loads Metro alternative way and keeps it in its place.

Restored Windows 7 Start Menu and Start Button There are tons of start menu clones (check the thread here), but they're all inferior. StartIsBack restores original Windows 7 start menu (with very special magic). It's NOT a replica, it's 100% original Start Menu as you remember it, with search, drag&drop, customization, etc. Start Menu is shown on Win key, as you expect. Start button is behaving as you remember it.

Disabled hot corners By default, all but charms hot corners are disabled. You can customize which corners to show on desktop. And this works better than in any start menu clones i've tried.

Modified Start Screen Now, we have Start and Start. Start Menu and Start Screen. Both search and start programs. Isn't that schizophrenic? So! Start Screen is no longer Start Screen. It's Apps Screen and shows only Metro programs (optionally). This way, Metro and Desktop experience are separated and Start screen is just a launcher for funny Metro apps, while real work and programs run in Desktop.