Well, SharePoint’s My Sites have a site-collection setting for ‘Portal Site Collection’. This setting modifies the Global Breadcrumb, and inserts another link there - so you can have a link back to your site-collection.

There a number of issues with that, though…

• Each user’s My Site is an individual site collection, so this value has to be set for each site, rather than just once. That’s not a problem, though - this can be done…
• …by feature stapling, which will allow you to set the ‘Portal Site Collection for any new sites. If you have My Sites in existance, that’s not going to work.
• And anyway, what if you’ve got more than one site-collection to link a user back to?

Thus, I found myself asking my colleagues, and they’d an interesting suggestion - adding new tabs to the My Site as if for a Personalization site. Thus, I went to my Shared Service Provider:

And I configured Personalization Site Links:

I set up the link:

And to my surprise, although it wasn’t a link to Personalization site, it still gave me a tab - which worked!

This is particularly useful as I could have tabs to serveral different Site collections.

… and a Document List content type, which had a couple of extra columns - % Complete, and Callback Number…

Next, I opened my site in SharePoint Designer, and went to the Data Source Library…

I clicked to create a new Linked Data Source and was presented with this dialog:

I clicked to Configure Linked Data Source… and I could now add my data sources…

… which I did…

And clicked next, where I was given a choice of type of linking to do…

I could merge the sources into one big list, or join them into wider records. I’ve never really done the latter, but I imagine that there are lots of uses for that - linking Order items with Customer items, and so on. Think of it as just like joining database tables in a SQL query. For this, though, I just merged my sources into one list.

This gave me a Data Source which I then used for my DataView …

Presto! Items merged, despite their having different columns. Ones that an item doesn’t have are just shown as blank.

• the Server
• possibly (but not necessarily) a managed path and site collection
• possibly (but not necessarily) a site
• possibly (but not necessarily) a folder (such as ‘/lists/’)
• possibly (but not necessarily) a list/library name
• possibly (but not necessarily) a folder in a Library
• the item itself.

Suffice to say, with all those optional bits, decomposing a URL to find the site is really hard. There is, however, a slightly obscure way of find this. You can create a site collection (SPSite) with a full URL, and then simply call OpenWeb() without any parameters to return you the site (SPWeb):

string path = "http://example/examplesite/_layouts/settings.aspx";
try
{
using (SPSite siteCollection = new SPSite(path))
{
using (SPWeb site = siteCollection.OpenWeb())
//Do something with the site
}
}
}

I found this when looking at the MSDN docs for SPSite.OpenWeb(). Check out the examples in there.

It’s a little weird that the SPSite object remembers information about how it was opened like that. But it is useful to know.

However, I think I’ve just found my first big gripe. Performing standard system maintenance on my desktop, which is used pretty much exclusively as a VM host, I noticed that my primary drive was using 40Gb. That seemed rather high for Vista, Office ‘07, SharePoint Designer and Adobe Reader. Naturally, I broke out Treesize:

So, 28% of my drive is being taken up by the winsxs folder. What the hell is that?  Well, I Googled - and Vista Club has a good description … 

In Vista, there isn’t a ‘dllcache’ folder and nor can you find the ‘i386′ folder, where the system caches (like in XP) all it’s source modules. It is this WinSxS folder which stores the shared components of side-by-side applications. These files can be multiple versions of the same assembly or application. Every side-by-side assembly has a unique identity. One of the attributes of the assembly identity is its version.

Okay, but why so big? Well, it carries on…

As mentioned, Windows stores the old dll’s and library components in the WinSxS folder. Now if a newer version of this file is a part of the OS, but a particular application requires a particular older version for running, then the older version from the winsxs folder will be used, leaving the newer version in its present place, for other applications which may require it.
 
So obviously you cannot delete this directory or move it elsewhere. Nor is it advisable to delete anything here, as such a step could probably make your applications un-workable or even break your system! If you have many applications installed, you can expect to have a jumbo sized winsxs folder.
 
If you delete components from the WinSxS folder like the manifets or the assemblies, etc, you could be in trouble.

….

Compressing the folder is also a no-no, as it could cause problems during WindowsUpdates or while installing a Hotfix.

Holy crap! It’s a 13Gb directory that I can’t tidy up, delete, move or compress! Who designed that, and were their brains on holiday? I get why they did it - to maintain backward compatibility - but sometimes, maybe it’s just best not to. I mean, VB.NET isn’t backwardly compatible with VB6. Office 2003 and 2007 files are only somewhat compatible. Win 95 wasn’t backwardly compatible with 3.1. The .NET framework 2 wasn’t backwardly compatible with 1.1. And in all those cases, it’s worked out for the better. Please Microsoft, break with backwardly compatibility and get rid of this folder.

It makes sense, I guess. You do need a master page which doesn’t run in an authenticated context, and a brief check shows that this is the master pages for the out-of-box Forms authentication page.

I decided to check out if there were other master pages in Layouts - and there were. My list of master pages is now:

1. Default.master - used in normal sites. Often replaced for publishing sites.
2. MwsDefault.master - use in Meeting Workspaces. Awkward to replace, but can be done through code or SharePoint designer.
3. Application.master - used in ‘administration’ pages, and sometimes pages such as ‘View all site contents’, Recycle bin, and uploading a file.
4. Simple.master - used in pages that need to run in an unauthenticated context - such as error pages, or forms authentication pages.
5. sspadmin.master - seems to be used editing audiences, in excel services configuration, in profile administration, and handful of other places.
6. layouts.master - no idea what this is for. Can’t find where it is used!
7. dialog.master - used in the little pop-up dialogs, such as when you’re adding or editing a Hyperlink.
8. pickerdialog.master - used in pop-up dialogs such as the web part gallery, or selecting users.

The master pages in italics are comparatively easy to replace; the rest are much more difficult. And try as I might, I couldn’t change the master page for Errors in my HTTP Module. I guess what I’ll do is write another one to redirect errors to my own custom-branded error page.

It worked - but the links to the announcements themselves didn’t work. The short summary is that is that it appears RSS feeds from Content Query Web Parts don’t work when you query for certain types of list.

What I realised was that there was no server in the URL, so my feed reader’s link to the item was going to a URL starting:

http://_layouts/CopyUtil.aspx?Use=id&Action=dispform ...

I’m now tempted to register the _layouts domain ;p

Anyway, checking in the RSS source showed links to incomplete URLs:

Hmm. So how to fix it? I did some Googling and found references to similar URLs, and the solution seemed to be to modify the ContentQueryMain.xslt file (found in the Style Library > XSL Style Sheets). I couldn’t see why this should make a difference - it’s not to do with the RSS feed - but I thought I’d try just incase I’d misunderstood.

I hadn’t. It didn’t work.

I then wondered if the RSS feed worked for a new Content Query Web Part that had just been added to a page. I added one, and enabled the RSS feed. It did work - the URLs in the links contained the server.

Curious.

I exported both web parts as XML and differenced them. One thing I noticed was that the UseCopyUtil setting was false in the working web part, and true in the one that wasn’t. I changed this to true, imported the web part, and now the URLs had a server at the start. They weren’t valid URLs still, so I swapped back.

Next, I wondered what would happen if I took the web part with the working RSS feed, and slowly changed it over to the query I wanted to run. Presumably, at some point it would break.

It did. When I selected that I wanted to query for Items in an Announcement list rather than a Pages library. I changed my query back to Pages - and the RSS feed worked. I tried changing to query for Document libraries - and the RSS feed worked. Then I tried changing the query to get items from Discussions lists - and the RSS feed failed again.

I then took a look at the page being used for the RSS feed (feed.aspx). All it was really doing was inheriting from the Microsoft.SharePoint.Publishing.Internal.CodeBehind.FeedPage page class. I broke out Reflector, and tried following what was going on inside, but to be honest, it was too complicated, and I couldn’t be bothered.

Either way, it looks like the Content Query Web Part’s RSS feed doesn’t work if you’re querying for Announcement lists.

This default behavior can be changed on a per web application basis to return only enabled user accounts. In the example below I have configured this for the web application http://moss using stsadm. This command configures MOSS 2007 to use a custom LDAP query.

stsadm -o setproperty -pn peoplepicker-searchadcustomfilter -url http://moss -pv (!userAccountControl=514)

If you are interested in putting together more granular filters I strongly recommend the following guide on LDAP, LDAP Query Basics - http://technet.microsoft.com/en-us/library/aa996205(EXCHG.65).aspx

Nice one Brendan

• Themes allow you to provide a bunch of CSS styles. So does the AlternateCssURL.
• Themes have to be applied site-by-site (unless you do some programming, or extend STSADM). Alternate CSS or Master Pages can be changed for a site and subsites, for a Publishing site anyway (although with some limitations).
• Alternate CSS can be provisioned from a single URL - so benefiting from browser caching just like images.
• The Alternate CSS can style the DatePicker. Themes can’t.
• The Alternate CSS can style Application Pages, just like Themes (but unlike Master Pages).
• No problems with ‘Style Merging’ when the Theme is provisioned.
• Style can be updated in one location.
• Themes can be changed through the UI for any site. Only Publishing sites have a user interface for changing the Alternate CSS.

Yes, many of the same distinctions between Master Pages and Themes still exist, and the same sort of analysis and choice of approach should be done. However, it means that we can kind of ignore Master Pages, and just focus on Alternate CSS vs Themes - and it seems to me that apart from the last point above, the Alternate CSS approach seems to equal or better than Themes.

Regarding the last point, well, to apply my theme I’d set the Alternate CSS in a Feature Receiver. This is okay, though, as I’d probably be writing one anyway to apply the correct master page to different types of site.

So, the answer I’ve come to now - I’d use the _layouts directory - create a sub-directory for your brand, and put your CSS and images in there.

The code on the page looked like:

<td id="_invisibleIfEmpty" name="_invisibleIfEmpty" valign="top" height="100%"> <WebPartPages:WebPartZone runat="server" Title="Left" ID="CenterLeftColumn"><ZoneTemplate></ZoneTemplate></WebPartPages:WebPartZone> </td>

…and there were a lot of table cells looking like that, and with that id. Lower in the page I then noticed:

<script language="javascript">if(typeof(MSOLayout_MakeInvisibleIfEmpty) == "function") {MSOLayout_MakeInvisibleIfEmpty();}</script>

Hmm. So if this function exists, it’s run. For it’s name, it sounds related. After some digging, I found it in defined in:

• IE50UP.js
• IE55UP.js
• NON_IE.js

Interesting… IE5 is, well, pretty ancient now, so I’m guessing that this stuff is a bit, well, WSS2 legacy. What does the code do? Well, looking in one of those files:

function MSOLayout_MakeInvisibleIfEmpty()
{
var allElements=document.getElementsByName("_invisibleIfEmpty");
var agt=navigator.userAgent.toLowerCase();
var isNav=((agt.indexOf('mozilla')!=-1)&&((agt.indexOf('spoofer')==-1) && (agt.indexOf('compatible')==-1)));
var isIE=(agt.indexOf("msie")!=-1);
for (var curElement=0; curElement < allElements.length; curElement++)
{
if ((isIE && allElements[curElement].childNodes.length==0)
|| (isNav && allElements[curElement].childNodes.length <=1))
{
allElements[curElement].style.display=”none”;
}
}
}

Yup, so it makes table cells called _invisibleIfEmpty not display if they have no child nodes. That’s a little strange though - table cells do collapse anyway if there are no contents, and doing this in Javascript just seems a bit old school.

I went and looked inside the Splash page layout and found that that is exactly how it works - the table cell’s borders don’t appear if the cell is empty:

However, the table cells do still appear to take up space - ‘cos each table cell has a width set:

If we took that width=”50%” away, though, the cells would just collapse as normal for empty table cells in HTML.

Still, the who thing did make me wonder - could I do my ’round cornered borders on web part zones’ with Javascript?

I checked my style, and there it was. So I tried adding a margin, or padding in the cell that contained it.

Wrong. It needs it’s height to be set to 100%.

.ms-globalbreadcrumb .ms-HoverCellActive,
.ms-globalbreadcrumb .ms-HoverCellInctive,
.ms-globalbreadcrumb .ms-HoverCellActiveDark{
height:100%;
padding-top:1px;
}


They’re just views that happen to use date columns…

So you can configure the relationship between your list columns and the views…

That’s an interesting concept. We can use this with any list! For example, the start and end dates for help desk calls, or for displaying when things happened. Neat! I can’t think why I didn’t realise this before!

Go to Settings > Recent Comments and hit Update Recent Comment Options. ‘Nuff said.

And select one:

Your site will then have a different theme. Great! But how does this work? Well, lets open our site in SharePoint Designer:

That’s right, a folder _themes, containing the theme we just applied in a folder. That theme is the CSS and all the images. Because of SharePoint’s virtual file system, that isn’t as inefficient as it sounds, but ‘cos our theme is being provided from our site, each CSS file and each image will be at a different URL for every site. This might be fine if you wanted to be able to  customize the theme on a per-site basis (and you can!) - but it eliminates some of the advantage browser caching, and I don’t think most organisations will want their theme customised on any sites - branding should be consistent.

The observant of you will notice multiple CSS files in the Theme folder (’Plastic’ above). How does that work?

Well, if you look in 12/TEMPLATES/THEMES, the themes themeselves don’t have the CSS file that you’ll find referenced in your Site’s pages. Instead, when you apply the theme, SharePoint takes all of the CSS files and appends them together into one file which it uses (Plas1011-650001.css, in yellow above).

I only actually found this as most instructions on creating a new theme start by telling you to copy an existing one. This is fine, but I forgot that there were more CSS files than just my Theme.css. Specifically, the mossExtension.css file was being added after the end of my themes, and it was overriding some of the styles I’d defined. As the order of the themes being defined is very important, I do think this ‘adding together’ of CSS files is somewhat dangerous.

Anyway, once a single CSS file has been created, the master page will insert a URL to that file into the page <head> section, and that’s how your styles apply.

I’m with Mads on this - to me the killer feature of SharePoint is how it works with Offices, and template functionality is key in that. Demos of document properties, quick parts and then the list columns in Sharepoint have a very high wow-factor (rightly so - I think it’s pretty neat too!)

However, templates are something rarely used in my experience, or not use properly anyway. Often it’s seen as an unnecessary effort, and usually everyone just starts creating their own documents from blank, or deleteing the contents from an existing document and using that one, or copying and pasting across - none of which are pretty scenarios.

Then again, often the guys actually making these templates need a good course of ‘How to use Word’. No so much with Excel - I think that people accept you need a bit of training to use it - but Word seems to be an issue.

Anyway, the point is, do try and plan building templates into your project. And that there is probably a consultancy opportunity in trying to generate these…

Well, you can activate a new master page in a Feature Receiver. And we can detect the type of site we’re dealing with. Why not combine the two techniques:

private const string MASTERPAGEFILE = "MasterPageFile";
private const string MEETINGMASTERPAGEFILE = "MeetingMasterPageFile";

public override void FeatureActivated(SPFeatureReceiverProperties properties)
{
SPWeb site = properties.Feature.Parent as SPWeb;
SPFeatureProperty masterFile = properties.Definition.Properties[MASTERPAGEFILE];
SPFeatureProperty meetingMasterFile = properties.Definition.Properties[MEETINGMASTERPAGEFILE];
char[] slashes = { ‘/’ };
//Is the site a Meeting Workspace?
if (site.WebTemplate.StartsWith(”MPS”))
{
site.CustomMasterUrl = site.ServerRelativeUrl.TrimEnd(slashes) + “/_catalogs/masterpage/” + meetingMasterFile.Value;
site.MasterUrl = site.ServerRelativeUrl.TrimEnd(slashes) + “/_catalogs/masterpage/” + masterFile.Value;
}
else
{
site.CustomMasterUrl = site.ServerRelativeUrl.TrimEnd(slashes) + “/_catalogs/masterpage/” + masterFile.Value;
site.MasterUrl = site.ServerRelativeUrl.TrimEnd(slashes) + “/_catalogs/masterpage/” + masterFile.Value;
}
site.Update();
}

• Any type of site will let you change the theme.
• They style the Site Admin pages too.

…but there are some pain points too:

• You can’t apply a Theme to a site and all subsites - well, not unless you write you own code to apply them programmically, or do something like extend STSADM.
• You have to use feature stapling to automatically apply a Theme on site creation.

It’s certainly worth considering how your theme will be applied up front - although you could using the Alternate CSS url to ‘apply’ your theme. Personally, I think I like the idea of using a Feature.

I’d been working on my nice little brand, and on the item creation pages it’s supposed to look something like this:

Nothing strange or startling, just some new colours. However, my customer found that if he moved his mouse over the links in the breadcrumb, he saw weirdness:

For some reason, on mouse over he was seeing bits of the page redrawn as background. WTF? Clearly, something about my styles was breaking the rendering in IE7. I know I don’t have any background graphics like what I was seeing - it has to be bits of the page redrawn.

I opened up the IE Dev toolbar, and took a look. None of the styles in use were incorrect - their code was fine. I started to take the scientific approach of, um, removing the styles that were in use one-by-one from the branding to see when it started working. It didn’t. Uh, oh. Out of panic, I just started removing all of the styles related to the PageAreaTitleFrame and it’s contents. When I removed .ms-pagetitleareaframe table it suddenly started working. But I could see that this wasn’t used - the IE Dev toolbar was showing me this.

WTF? The code for the style itself wasn’t complex, either! How could this break the rendering?

In particular, it appeared to be the line setting the background-color to white. How the hell does setting a background-color of a style that isn’t even being used in the page break the rendering of IE7?

One of my colleagues suggested looking for the ms-pagetitleareaframe in the source. He doesn’t love the IE Dev toolbar as I do, but to humour him, I took a look:

What? There it is! And it’s on the table cell that encompasses the page title area. But looking at the IE Dev toolbar, that cell is styled as ms-areaseparator.

I wondered if the problem was related to the cell’s ID, so I searched Google for onetidpagetitleareaframe - and the first hit I got was ‘Be careful about hard coding Javascript in SharePoint 2007′ (Good post, worth a look). Interesting thought - could Javascript be causing the problem? I tried searching for onetidpagetitleareaframe in my page source, and right at the bottom of the page I found:

<script id="onetidPageTitleAreaFrameScript">
document.getElementById("onetidPageTitleAreaFrame").className="ms-areaseparator";
</script>

The air turned blue.

So, my problem was a result of Javascript and my background colour for a style that was used for a page initially (but was replaced by Javascript as the page load ended) interacting and breaking the IE7 renderer. But not on the page load, but on the mouse moving over it. Now if that isn’t a knotty and specific set of circumstances, I don’t know what is!

My solution - I redesigned my styles so I wasn’t setting the background colour….