Marius's Personal site

In can you don't already know it, Windows 7 supports booting from VHD, which I'm a big fan of. This way you don't need to worry with partitioning scheme, shuffling around when you want to install additional Windows versions, backup / restore or these kind of issues.  Booting from VHD allows you easy migration to Hyper-V if you chose to do so in the future.

However if you choose to install your OS to a VHD then you need to know that you will loose some things:

1. Hibernate support. Unfortunately it seems that there is no way (at least not in the RTM version) to enable that. Sleep support still works fine, but if you need hibernate (or hybrid sleep) then you need to install to a partition.
2. Calculate System Score. WinSAT fails to calculate the disk score if your system boots from VHD, because it wouldn't be accurate. Fortunately there is a solution for this. All you need to do is to launch an elevated command prompt and type the following command:
   winsat formal -drive D
   I found this while trying to setup an HTPC that needed to run Digital Cable Advisor, and while trying to calculate the system performance score it would crash very badly. It seems that they forgot to check if WinSAT actually completed successfully before parsing the results!

Hope it helps!

And if you find a way to enable hibernate while booting from VHD just leave a comment.

A few days ago Red Gate has announced that it acquired Reflector from Lutz Roeder, which is big news given the fact it is kind of core tool for .NET.

In case you don't know yet Reflector was listed in top 10 of the tools that every .NET developer must have, as can unlock many of the misteries of the .NET Framework and almost every other tool based on .NET. Red Gate announced that it will keep Reflector free so we don't need to worry at least for now and maybe we will find more addins for Reflector and community will be more interested to create for it.

In my opinion this move was somehow determined by the fact that some of the products from Red Gate's portfolio have now some competition from Visual Studio Team System, but this is the subject of another post.

One of the first thing that I want to emphasize is that Sharepoint is built using ASP.NET. What this means for a developer? You can do all the things you can do in ASP.NET, and more! FindControl for example, you can enumerate the list of controls from code behind, Master pages, you name it.

One of the hidden gems of Sharepoint (at least for developers that want to extend the platform) is the Delegate control. Delegate Control works similar to delegates in .NET and basically are placeholders where Sharepoint platform instantiate specific controls depending on the installed features. So for example you can provide an alternate implementation of the search box control to display your bells and whistles instead of the default one, which is a great feature if you want to override something you don't like. Now comes the better part: Delegate control has another mode where the platform instantiate all the controls with the same id no matter what Sequence number they have, basically allowing adding of generic functionality, without having to implement something you don't want to.

Example: you want to display your custom magical text in all the pages of your Sharepoint site, no matter what. If you would override SmallSearchInputBox control, then you would be able to display the text, and also you would have to implement the search functionality also. This mode is called AllowMultipleControls and is placed as an attribute for DelegateControl tag. Now everything is fine, however how do you find such control? Sharepoint was kind enought to provide AdditionalPageHead delegate control in the master page, basically allowing to be able to intercept every single request for that farm or site. In your OnInit/OnLoad event you can filter the request based on your needs and then inject your own control tree, or alter the request as needed.

To do that you would create a feature like: .codeblock, .codeblock pre { font-size: x-small; color: black; font-family: Consolas, "Courier New", Courier, Monospace; background-color: #dddddd; } .codeblock pre { margin: 0em; } .codeblock .kwrd { color: #0000ff; } .codeblock .str { color: #006080; } .codeblock .html { color: #800000; } .codeblock .attr { color: #ff0000; } .codeblock .lnum { color: #606060; }

   Feature.xml
<?xml version="1.0" encoding="utf-8"?>
<Feature Id="INVALID!-GUID-0000-0000-000000000000" 
    Title="My Feature Title"
    Description="My Description"
    Version="1.0.0.0"
    Scope="Web"
    xmlns="http://schemas.microsoft.com/sharepoint/">
    <ElementManifests>
        <ElementManifest Location="Controls\mycontrol.xml" />
    </ElementManifests>
</Feature>

   Controls\mycontrol.xml
<?xml version="1.0" encoding="utf-8"?>
<Elements xmlns="http://schemas.microsoft.com/sharepoint/">
  <Control
      Id="AdditionalPageHead"
      Sequence="100"
      ControlClass="My.Own.Control.Class" 
	ControlAssembly="My.Assembly, Version=1.0.0.0, Culture=neutral, PublicKeyToken=................">
    </Control>
</Elements>

    My Code.cs
public sealed class My.Own.Control.Class : Control
{
    public My.Own.Control.Class()
    {
    }

    protected override void OnInit(EventArgs e)
    {
		base.OnInit(e);

		if (some conditions) 
		{
			LiteralControl ctl = new LiteralControl();
			ctl.InnerText="Hello everyone!"
			this.Controls.Add(ctl);
		}
    }
}

Also remember to place your control in SafeControls section of web.config like:

<SafeControl Assembly="My.Assembly"
 Namespace="My.Own.Control" TypeName="Class" Safe="True" /> 

This technique can be found useful for example if you want to alter the behavior of webpart pages and you don't have control over their base class (for example if they are in a webpart document library).

After running WINNT32.EXE in the working Windows 2000 installation, program asked for a reboot, I agreed and BOOM! blue screen. The famous BSOD stating stop 7F (aka INACCESIBLE_BOOT_DEVICE). Trying to debug the issue via WINDBG and serial cable (yeah, I had to run to COMPUSA, they are closing the stores in my region and get a plain old serial cable) didn't revealed anything terribly wrong. After long sessions of registry compare revealed that the culprit was VIAAGP1.SYS driver. If you are wondering what has to do VIAAGP1 with disk access (INACCESSIBLE_BOOT_DEVICE is directly tied with disk access), well, the only explanation I came up with is that VIAAGP1 is marked in UpperFilters key for the AGP bus which has PCI.SYS as a driver, and apparently VIAAGP1 failure was tearing down the entire PCI bus access.

If you think it's the end you are terribly wrong. After that, SETUP was crashing with access violation at the "Installing Devices" step. The culprit? Apparently, MS_NDISWANBH instance which is the network monitor filter driver for WAN.

And after a week of rebooting and debugging, Windows 2003 finally came up, with the fenomenal problem of not starting TCPIP driver. Reinstalling the protocol fixed the issue.

The short conclusion: now it works fine!

So if someone face similar problems this is what I've done (standard disclaimers apply, I'm not responsible if anything fails on your machine, make backups, you know the story):

• Delete VIAAGP1 from UpperFilters entry in HKLM\SYSTEM\ControlSetxxx\ENUM subtree. More specifically the AGP bus, but there is no standard location. Just search here for VIAAGP1
• Delete CDR4_2K entries UpperFIlters from HKLM\SYSTEM\ControlSetxxx\Control\Classes subtree
• Delete all the entries for NDISWANBH in HKLM\SYSTEM\ControlSetxxx. This might was the cause for TCPIP driver not starting, however since setup kept crashing I could find a better solution at that moment. You might want to backup all the keys you are deleting, and restore them after the setup completes, YMMV

And a side note: kernel debugging via WinDBG / serial cable at 115200 bps is not that bad. However the best is not to need such a debug.

This post is inspired from another post (in Romanian, sorry for the not Romanian speakers), which asks about the usefulness of System.Reflection.Assembly class.

This class is primarly your one stop shop if you want to implement a plugin host in .NET Framework. The implementation of a plugin host in .NET Framework is somehow different than the old "C/C++" way of implementing by the simple fact that we no longer have header files to ship to the plugin writers in order to describe our interface. But no worry, this doesn't means that it's more complicated to do such a thing in .NET. The closest way to do this is to generate a class library DLL project, that doesn't implement code (or maybe just some utility classes) but to have the definitions of our interfaces. This is the easy way, but in my opinion it's not the cleanest way. If we go this route we will have to ship our "interface" DLL to clients since it's a required dependency. For large projects where you have hundreds of files shipped to client, it might be fine, but for small project where you ship just your host and few plugins the clutter it's less desirable. A clever way to do this is to implement your interfaces in your host project, and make the plugins dependant on your host .exe file. This way you will have always up-to-date dependencies, since you cannot break the host (or, your code won't compile at all). Prior to Visual Studio 2005, you weren't able to add to a project a dependency to another project whose output is a .exe file. This was recognized more or less a bug, since .exe and .dll files doesn't differ at all internally, so the workaround for Visual Studio 2002 and 2003, was to build your project, get the output, rename it to .dll extension, and add the dependency to your plugin project.

If you are new to plugins in .NET and you come from C/C++ plugins a warning that might save you a lot of time: do not try to reimplement your host interfaces with the same definition in your plugin project, just to remove the intermediary DLL file dependency (the easy way above). IT WON'T WORK. Objects in .NET (interfaces, classes, etc) have the same identity only if they have the same name AND come from the same assembly.

Someone said there that is not normal that Visual Studio accepts dependecies to .EXE files. Well, the fact that they prevented explicitly this (I believe, since they were filtering some types of projects) in VS 2002 and 2003, and now it works in 2005, it might mean that some people finds the feature useful. And I come and say why not accept dependencies of .EXE files/projects?

Have you seen the design of Vista and Office boxes? Apparently I'm not the only one that after seeing the Vista and Office 2007 boxes thought that they look like the Office for Mac retail boxes.

In case you want to know more about Vista, you can use the "Windows Vista blog" to find out, as it's packed with a lot of useful information.

The main problem with Microsoft SQL Reporting Services (either 2000 or 2005) is the lack of capability of uploading or downloading multiple RDL files at once. I know that the RS.exe utility exists, and it's presence is more proeminent in the 2005 version by adding the SQL Management Studio ability to script files of RS.exe, but you cannot upload quickly a locally stored folder full with RDL files, or for example copy a lot of reports from one server to another. That's why I decided to create this utility named RSUploader at this time at the version 0.5. Although I intent to update it a little bit I cannot promise that this will happen or a timeline for it's improvement.

RSUploader was built using .NET framework and presently the user interface is not the best in the world, but still gets the job done. The minimum requirement is .NET Framework 2.0 and it can connect to Microsoft SQL Reporting Services versions 2000 and 2005. You can download it from here

Among it's features you can:

• Browse the folder structure of Reporting Services
• Download all report definitions from one Reporting Services folder
• Upload a local folder containing RDL files in a reporting services folder
• Delete reports
• Hide/Unhide reports

In order to successfully use RSUploader you have to type the URL of Reporting Service Web service, usually http://[machine]/ReportServer, do not swap it with the URL of Reporting Service Manager, and also if you are not using Integrated Authentication, type the username and password in the corresponding textboxes, and click Refresh in order to connect to the Reporting Services. Any time you change the URL or user / password you must click on Refresh in order to pick the new location.

Then expand the folder structure in the left view and in the right view you will have the actual list of reports and datasources. The checkbox in the front denotes the visibility of the item (if it is not checked it means it is hidden). Also you will have fix datasources for reports if you uploaded reports before you created your datasource and the GUID that Reporting Services is using for matching the datasources is not in sync.

If you find this utility useful please drop a note using the contact form.

Of course, the information in this post and the utility is provided "AS IS" with no warranties, and confers no rights. I am not responsible if any thing doesn't work as expected or for any loss of data.

As I was saying quite some time ago I decided to upgrading my router firmware. Originally, I wanted to buy this particular model, due it's upgradeability to a open source firmware. The upgrade was a little bit postponed due the fact my computer decided to stop working so I finally managed to do it. If you ask me which were the primary reasons of the upgrade, well static DHCP allocation and SNMP daemon are in the top list. If Linksys would add these feature in the stock firmware, right now I would not add this post. Static DHCP allocation is a very nice feature these days, as you can reshuffle your network (I mean especially in a more business scenario) without touching all the computers/devices out there, but eliminates the disadvantages of changing the device address without any reasons, or not having a predictable numbering scheme. As I said these are quite important in a business scenario, but even at home if you have more than a few devices it worths. Actually, I had a problem with the stock firmware that if I unplugged the router all the devices would have new addresses that usually are not the same, depending which device was faster to request addresses.

The next feature I do not think requires much explanation: graphing the network traffic is quite essential in pinpointing potential problems, so here it goes.

And the lucky winner was DD-WRT after all. I investigated a little bit over OpenWRT, but I decided to go after DD-WRT, since it seemed to me a little bit more interesting. I agree that both distribution have pretty much the same offerings, and even found that they even want to merge to concentrate the effort.

The upgrade process was plain simple, just followed the Wiki installation tutorial and in about 10 minutes it was completed. I decided to skip the mini version and go directly to the voip version. I found that works perfectly going directly to the bigger version, with the warning that it might not work for some devices/versions, but apparently if it doesn't work it doesn't brick your router. This being said I very happy with my latest DD-WRT v23 SP2, and it seems very stable to me. So, keep up the good work with DD-WRT!