Google And Microsoft : Twin Effect

The evidence can be found everywhere. Headlines, headways and in some people's head.
The pattern is recognizable given that one have been around since DOS days.
Google, the company that was once just associated with "search" is no longer staying where its domain is. It has tentacles that reaches far and wide but with one clear difference. The community don't realize it or maybe they do but they sure don't hate it.

Perhaps MS's reputation and brand was so powerfully seared into the hearts and minds of computing folks that we could only identify MS as the only possible gigantic sofware company that we could hate. One that squashes its competitor mercilessly using various "partnership" tactics through the history of IT. Novell...Nescape....scandisk....C++ builder...Word Perfect...Quattro Pro....OS/2......Isn't scandisk just a program, yeah, so was Netscape browser. Sure some of these companies survived but became obscure tech names, a fraction of what they could have been.

Times are changing indeed. Windows Mobile is losing out. Microsoft Live Search is a joke. A victim of its own game in an ever changing world. Its rare to hear MS losing in an area that it plans to focus or profit from but the new kid in town is nothing like MS have seen. Previously everyone had to fight MS in MS own turf...and since that turf is the OS itself, it was rarely a fair fight. Its 90% of the World's desktops dear, get real. But what if you challenge MS in a very unlikely place...the cloud?

When MS IE7 could not open the Google Wave page and Chrome could use it perfectly (ok firefox and other browsers worked somewhat) , you get a very familiar feeling. This creepy Dejavu feeling just jumps out of the page into your lap.

Android phones? Google Doc? GMail for Corporates? GAE hosted domain.

China always had a long history of bad Emperors and those who attempt to topple them. Given such power and position, it was usual for Emperors to behave like morons and do what they want regardless of the citizens sentiments. Heck an Emperor gets to have 600++ wives!

But Chinese History also teaches us one thing, topple one Emperor, the one taking over his place is no better.

Google Wave Hackathon Malaysia

Today i attended the Wave hackathon. Some lucky fellas here are supposed to be the first in Malaysia to get google wave accounts (for the sandbox) .

Nazrul (the founder of MYGTUG) gave an overview and everyone quickly surfed and googled for more info . Basically for most of us, its the first time we are exposed to this Wave thing. This whole idea of blurring the line between emails and chats is boggling but opens up a whole new way of using "emails".

Due to some problems faced by other participants, somehow my "robot" was the only one demo-able at the end of the day. I guess my decision to just use the robot for fun things turns out to be easier to implement, Java programmers also complained about some issues with the api and Wave's unstability in handling certain events.

I went in with Python and i was using WingIde Pro, that must have been a real time saver since the autocompletion and module search gave lots of revelation of how some of the Wave Api works and fixes most easy errors. The problem is that you can't do a local test, so all the code must be uploaded before you can test it , simple errors will be a waste of time.

Here are some "traps" to avoid in creating a "robot" :-

1. If you add a capability to the robot AFTER you already uploaded your first robot, you will need to bump the version in your "whateverrobot.py" under the _main_ so that Google Wave will recognize that you are supporting a new event handling.

Java programmers will need to modify the capabilities.xml file to bump the version.

2. Don't just use the Tutorial example in creating Blips, you will end up modifying the blip

that the user posted instead of creating a new one since that is what the Tutorial example does.

You will get something like "User and robot" in the blip which is probably not what you want, instead just : newblip = root_wavelet.CreateBlip().GetDocument()

You can also try blip.CreateChild() if want it to appear below the original blip.

You must use the Tutorial example to learn and do the first upload though! Just don't

copy everyline when u want to do your actual one.

3. Upload the tutorial example so make sure all your other setup is already working.

You will need to setup a Google Application Engine account and a choice of either

Python or Java. (for Python i was using ver 2.5.4). Setting up Google Application Engine requires a valid and real mobile phone (it will sms a validation code).

4. IE.7 is unsupported!! Yeah, i guess Google is doing to Microsoft what Microsoft used

to do to some companies. So make sure you just stick to Chrome when participating.

Overall, it was a pleasant and enjoyable event.

Here is the robot i wrote (for those who have google wave account) :

Id : 369robot@appspot.com

Have Fun!

Too much support?

I got this screen this morning when searching for something in MS.

Did MS just redirected all support calls to online search? ;-)

C# calling C DLL routines (YET ANOTHER ONE)

I am totally baffled on the examples given by sites on how to call C routines in DLL via C#.
Not only are many of them repeating one same example after another, you soon realize they are not very helpful when you want to do something that is common like passing a structure with all kinds of types inside.

So here is one example that serve as an answer to MANY of the pinvoke/calling C apis problem :- ( i wanted to say 90% but i can't justify or prove it ;-))

I got this example structure that has :

struct ToughCallDef
{
char * ptrName; // this will contain some return string values
int ptrNameLen; // normal integer
MEMBLOCK somestruct; // a structure
ANOTERBLOCK* structpointer; // structure pointer
int* ptrtoMyInt; // integer pointer
char * ptrID // this must be supplied
};

Just translate it this way :-

[StructLayout(LayoutKind.Sequential)]
public class ToughCallDef
{
IntPtr ptrName;
MEMBLOCK somestruct; // you need to define MEMBLOCK also
IntPtr structPointer;
IntPtr ptrtoMyInt;
IntPtr ptrID;
}

And if the call requires a reference to this "ToughCall" :

[DllImport("anotherc.dll")]
public static extern int Call_C_API(ref ptrToughCall);

define it this way instead :

[DllImport("anotherc.dll")]
public static extern int Call_C_API(IntPtr ptrToughCall);

Basically anything to do with Pointers, regardless of whether its integer *, char *, structure reference etc, just use IntPtr.


1 . If the member is a "char *" that you expect return values you point the IntPtr this way :

szReason = new StringBuilder(100);
zReason.Append(' ', 100); // empty string with spaces
ToughCall.PtrName = Marshal.StringToHGlobalAnsi(m_szReason.ToString());

2. If the member is a "char *" that you need to assign a value before passing in :

ToughCall.ptrID = Marshal.StringToHGlobalAnsi(StringVar);


3. If the member is a structure pointer, you assign it this way :

ANOTHERBLOCK o = new ANOTHERBLOCK();
ToughCall.structPointer = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(ANOTHERBLOCK)));
Marshal.StructureToPtr(o, ToughCall.structPointer, true);


4. After calling the C routine you will need to do the following to get the values :-

1. If its a string, use :
   string s = Marshal.PtrToStringAnsi(ToughCall.PtrName)
2. If its a structure use :
   Marshal.PtrToStructure(ToughCall.structPointer, o);
3. Free any allocated strings/struct pointers via :
   Marshal.FreeHGlobal

That's it.

Twitter (python api) on Nokia S60

I wanted a system whereby people can sms to a phone that will log it to twitter.
With this system an organization can keep track of what its support/sales staff are at any given point of time.
After tinkering around with the twitter python api and Nokia s60 python (1.9.6), i finally
got it working...(duh...that took me more than an hour...)

There is however ONE big catch (read end of the section)

Here is the code :

Gotchas :-

1 - Ah yes you also need to change that getusername routine in twitter, else it wont work

2- Pay attention to that pesky pyS60 Application Packager (READ the README file) , this is not py2exe. Basically, you will need to rename your main file to some fix default.py.

3- It will keep asking you for connection (access point) on each msg , unless your using some newer phones like nokia 5800 XM. If you are using N82, ALL the popular examples shown in the internet like these :

http://discussion.forum.nokia.com/forum/showthread.php?t=163939

or

http://snippets.dzone.com/tag/pys60

or

http://croozeus.com/blogs/?p=836

DOESN'T Work. The reason is due to some "compatibility problem" between socket and btsocket.

"Twitter" uses urllib2 not urllib, and these people probably have never tested it on that api. Someone from Silicon Valley once said that Nokia know nuts about promoting development, and thus he rejected their offer to develop for symbian and instead move the whole team to iphone.

In some ways, i agree....but i still love my nokias ;-) I reckon the same thing would have worked in Iphone much earlier and with less blood on my desk.

** Update :

Marcelo Barros from Croozeus :

Yes, that is the problem. I reported it at maemo some time ago.
I suggest you to use my twitter api. Extend it if necessary.
http://code.google.com/p/wordmobi/source/browse/trunk/wordmobi/src/s60twitter.py
It uses urllib and simplejson (I ported it to S60, code in the same dir).

Note : If you still want to use the twitterapi as it is and not the ported version, you will

need to hardcode the "getusername" and stick to "socket" only calls, avoid using btsocket.

C# calling Python scripts and processing output

Occasionally there would be a situation where the C# needs to call a python script and work with its output. While this might sounds like a piece of cake, it does have its gotcha, that is when
you want to process each line one by one and you happened to have somekind of timing (time.sleep) or some other python codes that works correctly under console but unable to be processed by C#.

Below is how you do it : (read until the end, there is a gotcha)

private static void OutputHandler(object sendingProcess, DataReceivedEventArgs outLine)
{
if (outLine.Data != null)
Console.Out.WriteLine(outLine.Data.ToString());
}

static void Main(string[] args)
{
Process p = new Process();
p.StartInfo.FileName = "python.exe";
p.StartInfo.Arguments = "c:\\test\\test.py";
p.StartInfo.UseShellExecute = false;
p.StartInfo.CreateNoWindow = true;
p.StartInfo.RedirectStandardOutput = true;
p.OutputDataReceived += new DataReceivedEventHandler(OutputHandler);
p.Start();
p.BeginOutputReadLine();
p.WaitForExit();
p.Close();
}


Assuming ur python script is the following (test.py) :

import time
import sys

for x in range(100) :
print x
time.sleep(1)


You will be surprised that you dont get anything. For some reason, time.sleep(1) and possibly
some other python library calls would result in the output not being flushed. The solution
would be to "flush" it :

for x in range(100) :
print x
sys.stdout.flush()

Vmware and the Virtualization Gotcha's

There is this lab that has 4 different Win2k servers, each running a different application.
Server A - Mailserver
Server B - Intranet application with MS SQL db
Server C - Vpn server and backup
Server D - Test server

Along the way, the admin decides to put all these dated servers into a VM and slot it in to a high end Server from Dell. The whole porting process took around 2 weeks and when it was over,
everyone was happy with the new setup. No more additional switching cables and multiple monitors lyring around and the perceived energy savings cost was a bonus.

However one day, the mailserver began to feel very slow, Procexp (sysinternals) itself was myteriously taking up 45% cpu and more and even with the mailserver service stopped the pc still feels awkwardly slow. Rebooting didnt that mail server VM doesn't help either.

When i came in to help out in this scenario, the first thing i went thru was the list of VMs running in the server. One particular VM is taking up 26% of the CPU of the main server, however that should not be the reason why it would affect the mail server VM. Upon closer inspection however, i notice some native apps running on the test machine VM (26%) that is using up the VM's tcp/ip port very quickly.
It then became logically clear that this was the problem, pausing that VM immediately restored the other server's performance and that was like a 100% improvement.

I suspect the problem is because the main server is still just an OS with the normal limitation of the 65535 ports on a single ip and single network card. Since all the VMs runs on this machine , that test application was just blasting away the network resource, this turns the allocation of the real main server NDIS packets resources into an ugly situation where each of the VMs are queueing up to get its allocation.

Another case solve. Another Hamster Huey and the GooiKablooi award for virtualization.