• site.org/blogname1
• site.org/departments/blogname2
• site.org/departments/blogname3
• site.org/services/blogname3

The association between blog IDs and sub-directory paths is determined in wpmu-settings.php, but the code there knows nothing about nested paths. So a person planning to use WordPress MU as a CMS must either flatten his/her information architecture, or do some hacking.

Challenge: hacking WordPress MU to support arbitrary directory paths for each blog

As with my multi-domain hack, the following assumes that you’re using the vhost=no setting, that you have access to and know how to manipulate your MySQL, that you have control over your DNS and know how to use it, and that you know how to configure Apache or similar. You’d also be smart to turn off any object caching you may have running, at least until we’re done doing direct database manipulation. The following also assumes that your wp-config.php sets the DOMAIN_CURRENT_SITE and PATH_CURRENT_SITE constants — if you’ve done a fresh install recently, it probably does, or you can check my domain mapping hack.

Hack The Path Mapping

Right at the top of wpmu-settings.php you can see how it strips all but the base of the URL path, but rather than mod that file, we can take advantage of an obscure MU hack: sunrise.php, which gets executed after some important WordPress components like the database class get loaded and before wpmu-settings.php.

To use sunrise.php, create a PHP file at /wp-content/sunrise.php and set define('SUNRISE', TRUE); in your wp-config.php.

Here’s the sunrise.php code I’m using:

if( defined( 'DOMAIN_CURRENT_SITE' ) && defined( 'PATH_CURRENT_SITE' ) ) {
	$current_site->id = (defined( 'SITE_ID_CURRENT_SITE' ) ? constant('SITE_ID_CURRENT_SITE') : 1);
	$current_site->domain = $domain = DOMAIN_CURRENT_SITE;
	$current_site->path  = $path = PATH_CURRENT_SITE;
	if( defined( 'BLOGID_CURRENT_SITE' ) )
		$current_site->blog_id = BLOGID_CURRENT_SITE;

	$url = parse_url( $_SERVER['REQUEST_URI'], PHP_URL_PATH );

	$patharray = (array) explode( '/', trim( $url, '/' ));
	$blogsearch = '';
	if( count( $patharray )){
		foreach( $patharray as $pathpart ){
			$pathsearch .= '/'. $pathpart;
			$blogsearch .= $wpdb->prepare(" OR (domain = %s AND path = %s) ", $domain, $pathsearch .'/' );
		}
	}

	$current_blog = $wpdb->get_row( $wpdb->prepare("SELECT *, LENGTH( path ) as pathlen FROM $wpdb->blogs WHERE domain = %s AND path = '/'", $domain, $path) . $blogsearch .'ORDER BY pathlen DESC LIMIT 1');

	$blog_id = $current_blog->blog_id;
	$public  = $current_blog->public;
	$site_id = $current_blog->site_id;
	$current_site = sl_get_current_site_name( $current_site );
}

function sl_get_current_site_name( $current_site ) {
	global $wpdb;
	$current_site->site_name = wp_cache_get( $current_site->id . ':current_site_name', "site-options" );
	if ( !$current_site->site_name ) {
		$current_site->site_name = $wpdb->get_var( $wpdb->prepare( "SELECT meta_value FROM $wpdb->sitemeta WHERE site_id = %d AND meta_key = 'site_name'", $current_site->id ) );
		if( $current_site->site_name == null )
			$current_site->site_name = ucfirst( $current_site->domain );
		wp_cache_set( $current_site->id . ':current_site_name', $current_site->site_name, 'site-options');
	}
	return $current_site;
}

The first few lines of the code do pretty much the same as the start of the wpmu_current_site() function in wpmu-settings.php, but starting with line 8 it takes a big departure.

That’s where it splits the requested URL path like /path/to/blog/and/stuff/ into pieces and constructs an SQL query against the wp_blogs table to identify the correct blog to serve the request. The following example shows how:

SELECT *, LENGTH( path ) as pathlen
	 FROM wp_blogs
	 WHERE domain = 'domain.org' AND path = '/'"
	  	 OR (domain = 'domain.org' AND path = '/path/')
	 	 OR (domain = 'domain.org' AND path = '/path/to/')
	 	 OR (domain = 'domain.org' AND path = '/path/to/blog/')
	 	 OR (domain = 'domain.org' AND path = '/path/to/blog/and/')
	 	 OR (domain = 'domain.org' AND path = '/path/to/blog/and/stuff/')
	 ORDER BY pathlen DESC
	 LIMIT 1

Optimization note

Setting a maximum depth (and array_slice( $patharray, 0, $maxdepth )) would allow the query to be cached up to that depth. Otherwise, the query must be executed for every page load. The $maxdepth could either be set arbitrarily, or could be determined automatically based on the maximum path length of registered blogs.

Setting Up New Blogs

Once you’ve hacked the path mapping (and tested that it didn’t break your current site), you can add a new blog at a nested path.

Create a new blog in the MU blog admin.

Unfortunately, MU strips the slashes from the URL path you just tried to set.

The new blog you just tried to create, but with a very different path.

Fortunately, you can set the path correctly in the MU blog editor, and it won’t break the path when you save there.

Set the blog path in the MU blog editor, MU won't break it when you save it this time.

Once you create the new blog, try to load it in your browser. You’ll quickly notice the stylesheet is missing, though the blog works and functions properly.

Hack The .htaccess

WPMU uses the following .htaccess rewrite rule to properly direct requests for files on the real filesystem:

RewriteRule  ^([_0-9a-zA-Z-]+/)?(wp-.*) $2 [L]

Obviously, that rule won’t work for deep paths, so I’ve replaced it with this rule:

RewriteRule  ^(.+)?/(wp-.*) /$2 [L]

And with that, you should be done.

The following was my email response to a thread on the web4lib mail list:

Okay, it must be said: you’re all wrong[1].

I can understand that news of a librarian being fired/furloughed will raise our defenses, but that’s no excuse for giving up the considered and critical thinking that this occasion demands.

Consider this: the principle’s blog reveals a reasonable person actively trying to improve academic performance despite crushing economic conditions. The communications show a level of transparency many of us can only wish for.

Unfortunately, many on this mail list seem to have come to the conclusion that this library was a stellar, but unappreciated example of everything that libraries should be, capriciously closed by a principle who secretly wanted to see the football team shoving bookshelves around on a hot summer day.

Go ahead, mock the story (and so far we only have one story about this) for suggesting that the books have been “re-organized” by subject, but the fact remains that this community didn’t think their library was organized in a way that met its needs. This suggests that either (a) it wasn’t well organized, or (b) the librarian had failed to educate the users and develop the finding aids necessary to help the community use the library.

Nobody here is banning or burning books. Nobody is suggesting that libraries are irrelevant. Far from it: this story about the modernization of a library to make it a more significant part of students’ academic activity.

News that a member of our profession has been furloughed is sad. But, news that a principle is investing time, attention, and money in the library is good. News that those two stories are one in the same should make us ask critical questions about how we and our libraries are positioned to serve our community.

[1]: Everybody but Robert L. Balliot, whose message has so far been ignored.

The description to David McNicol’s URL Cache Plugin raises more questions than it answers:

Given a URL, the url_cache() function will attempt to download the file it represents and return a URL pointing to this locally cached version.

Where did he plan to use it? Does he envision the cache as an archive, or for performance? Why hasn’t it been updated since 2005?

It caught my interest because I’ve long been interested in a solution to link rot in my blog. A real “perma-permalink” would be very useful.

Bret Victor offers the above design suggestions (from 2006) to Amazon in the book search results display (he’s comparing to this). I didn’t discover them at the time, but many of them are still relevant now. Bret notes that Amazon’s display doesn’t do a good job of answering the questions a person has when searching for books: “What is the book about?” and “is it any good?”

Unfortunately, these questions are completely unaddressed by the information provided. To see relevant information, the user must click on each listing individually. That is, she must navigate by hand instead of by eye, and must use her memory to compare information across time instead of space.

The problem is that this graphic was designed as an index into a set of webpages, but is used as a catalog for comparing a set of books. The purpose of this graphic should not be to return a list of query matches, but to help the user learn about books related to her topic of interest.

The Hanzo Archives Wordpress plugin is something I’d be very excited to use. Ironically, it’s disappeared from the web (though the blog post hasn’t):

We’ve released a Wordpress Plugin which automatically archives anything you link to in your blog posts; it also adds a ‘perma-permalink’ for the archived version adjacent to each original link.

An Amazon Web Services case study put me on to Hanzo a while ago, and in May 2008 I actually spoke with Mark Middleton (the markm who posted the entry above). Mark revealed that community take-up on the plugin and other general purpose web archiving services was below expectations. The company has since refocused on legal matters (even their blog tag-line has changed to “web archiving for compliance and e-discovery”).

I wonder if, now that the number of people and companies that have been blogging for years has grown, there might be more of a market for such a service.

Above is Peter Linsley speaking about Google Image Search at SMX West in February, 2009.

Meanwhile, Stefan Juhl suggests some JavaScript to break your site out of the image search result pages:

Many Google image search users are quickly clicking on to the direct image URL and thereby not seeing the page with the image. Also, it seems that many of the users don’t hesitate to click back to the image SERPs when they don’t see the image “above the fold” – probably because of Google image search framing the page with the picture and thus making it almost too easy to do so.

<script language="JavaScript" type="text/javascript">
if (top.location != self.location) top.location.replace(self.location);
</script>

After adding this to the one of my websites I saw a quite big increase in pageviews. It turned out that the visitors happily continued to browse around the website.

Who doesn’t want to be an anarchist librarian? Or a bibliophian?

It turns out that there are a lot of differences between Google’s regular web crawler and the Google News crawler. And though very few of us will find our content included in Google News, it still seems like a good idea to make our content conform to their technical requirements. Here are a few of them:

• In order for our crawler to correctly gather your content, each article needs to link to a page dedicated solely to that article. We’re unable to index articles from news sections which consist of one long page rather than a series of links that lead to articles on individual pages.
• If your articles are located in a drop down box, we won’t be able to crawl them. Google News is unable to crawl articles only accessible through a drop down menu.
• Google News does not recognize or follow Flash, graphic/image or JavaScript links which link to articles. Our automated crawler is best able to crawl plain text HTML links.
• Google News doesn’t crawl articles in PDF format, although this content is included on Google Web Search. Our automated crawler is currently best able to crawl plain text HTML sites.

Facebook’s guide to sharing details some meta tags to make that sharing work better:

In order to make sure that the preview is always correctly populated, you should add the tags shown below to your html. An example news story could have the following:

<meta name="title" content="Smith hails 'unique' Wable legacy" />
<meta name="description" content="John Smith claims beautiful football is the main legacy of Akhil Wable's decade at the club. " />
<link rel="image_src" href="http://www.onjd.com/design05/images/PH2/WableAFC205.jpg" />

As shown, title contains the preview title, description contains the preview summary and image_src contains the preview image. Please make sure that none of the content fields contain any html markup because it will be stripped out. For consistency’s sake, please use the <meta> tag to provide text data for the preview, and the <link> tag for any source urls.

The title and summary tags are the minimum requirements for any preview, so make sure to include these two.

WiFi and college choice

• 90% of college students say Wi-Fi access is as essential to education as classrooms and computers
• 57% say they wouldn’t go to a college that doesn’t have free Wi-Fi
• 79% say that without Wi-Fi access, college would be a lot harder
• 60% agree that widely available Wi-Fi on campus is an indication that a school cares about its students

WiFi and where they use it

• 55% have connected from coffee shops and restaurants
• 47% from parks
• 24% from in their cars

WiFi in the classroom

• 55% have checked Facebook™ or MySpace™ and sent or received e-mail while using their laptop in class
• 47% have sent instant messages to a friend during class
• 44% used Wi-Fi to get a head start on an assignment before a class was finished

WiFi and linkbaiting statistics

• If forced to choose, 48% would give up beer before giving up Wi-Fi

Survey methodology: “In conjunction with the Wi-Fi Alliance, Wakefield Research surveyed 501 U.S. college students in September 2008. The sampling variation in this survey is plus or minus 4.3 percentage points.”

Richard Wallace’s Juice project (Javascript User Interface Componentised Extensions) is a “simple componentised framework constructed in Javascript to enable the sharing of Ajax Stye extensions to a web interface.”

WordPress or Scriblio users might do well to think about it as a way to put widgets on systems that don’t support widgets, though as Richard points out, “the framework is applicable to any environment which, via identifiers contained within a html page, needs to link to or embed external resources.”

I got a little excited when Shirley Lincicum wrote to the NGC4Lib mail list:

[O]ne of the most frustrating things for me about Next Generation Catalog systems as they currently exist is that they seem wholly focused on the user interface and can, in fact, actually hold libraries back from designing or implementing improved “back end” systems because of the dependencies introduced by the new “discovery layer” applications.

I was excited because almost two years ago I wrote something like this:

Libraries are good at sharing data, but we’ve done a poor job of taking advantage of the network and new technologies to reduce the costs of sharing. …We recognize now that our data is living and evolving, but synchronizing available record enhancements with individual collections remains costly and laborious.

Without efficient mechanisms to share improvements, the value to any one library of trying to share what local improvements or corrections they make is limited, preventing libraries from benefiting from the network in ways that open source software development has.

Some, however, have called Scriblio a “next generation catalog,” so I’m anxious to point out the following: The Scriblio MATC Project Final Report, which hits on some of the above points; this (not funded) IMLS proposal, which fully embraced the challenge (take a look at this diagram); and Scriblio 2.7 with its new internal data model, which finally delivers some of the answers I’ve had in mind.

On the other hand, I’ve resisted the label “next generation catalog” for Scriblio not only because the software does a pretty good job of hosting digital libraries in addition representing library catalogs, but because my hope is that Scriblio does more than put a pretty face on antiquated systems. It’s hard to deny the dramatic changes in writing, publishing, and information sharing, and WordPress is very near the center of it all (WordPress.com alone hosts about 150,000 new posts each day); I see an opportunity for libraries to participate at the start of information creation, rather than at the end.

Shirley continues:

If our metadata and communication standards, and the systems we use to manage the resources we collect, were open enough, and therefore able to be integrated seamlessly into general discovery interfaces like Google, Facebook, etc., it would allow librarians to focus on collecting and organizing stuff (which is challenging enough to do well), and let the folks with the resources to do really good usability research and hire lots of really good programmers to design the interfaces.

I might argue with the end of Shirley’s point, but the overall message that we build systems and data that integrates with Google, Facebook, and whatever else is next is a good one.

I should have done screencasts like the above long ago. It’s not that they’re great, but they are a wonderful excuse to use the canned lounge music I’ve got. Those videos are now on the front page of the official Scriblio site, and I did five more to demo the installation and configuration. Big thanks go to Collingswood NJ Public Library Director Brett Bonfield who let me use his library like this.

My slides for my presentation yesterday at code4lib are available both as a 2.7MB QuickTime and a 7.8 MB PDF, while the gist of talk went something like this:

Scriblio is an open source WordPress plugin that adds the ability to search, browse, and create structured data to the the popular blog/content management platform. And WordPress adds great ease of use, permalinks, comments/trackbacks/pingbacks, and other social and web-centric features to that structured data. But that’s not news. The news is that Scriblio now has an internal data model that supports much more sophisticated uses (slides 3 and 4). Whereas previous versions of Scriblio were mostly just display and social interaction interfaces to data that’s created or managed elsewhere, this new version supports soup to nuts creation and management of collections. Colby-Sawyer College’s archive (slide 5) is the first to implement this (take note of how the horizontal search layout makes the facets more visible and usable).

And that new data model also improves the usefulness of Scriblio to regular libraries (Collingswood (NJ) Public Library is shown on slide 6). Because Scriblio has an internal awareness of the metadata, it can automatically merge records from multiple sources (or multiple copies of the same record from the same source). The source of each piece of metadata in a record is identified and preserved (see the sourceid column in slides 7,8,9), allowing records to contain data from multiple sources (each with, perhaps, its own licensing terms). A practical example is enriching book records with data from LibraryThing’s Common Knowledge web service, making that data part of the index and facets in the local catalog, while also properly crediting the service when a record contains that data.

The automated merging of records enables a few new applications. Among them: the merging of an A to Z periodical list with the ILS’s inventory, or the creation of a union catalog from several systems. Slide 11 shows a prototype union catalog that shows materials (and their real-time availability) from three institutions in New Hampshire. Assembling that catalog was as easy as entering each ILS’s hostname and record number range in the harvester (slide 12).

I didn’t mention it during the presentation, but Scriblio is now built to work well in both regular WordPress as well as WordPress MU, the multi-user version of WordPress that allows a single installation to host many different sites (think WordPress.com) at a marginal cost to the hosting organization that approaches zero. The work to make Scriblio compatible with WordPress MU was supported in part by the National Endowment for the Humanities (there’s lots more to say about that project soon).

You’d think the top search results on the matter would be newer than 1999, but that’s where you’ll find this NYT article and PubLaw item story, both from precambrian times. Worse, both of those articles suggest that my links to them may not be entirely kosher.

The problem is probably that US courts have not spoken clearly on such a case. A Danish court in 2006 did, but I think that no case in the US has gone far enough to actually set a precedent. Another chance at settling this issue was lost earlier this month when BlockShopper settled, rather than continue a costly defense of such a case. The EFF is confident BlockShopper could have won, but that means little when the legal bills come in.

Related at EFF: Kelly v. Arriba Soft and Unlawful Internet Gambling Enforcement Act.

lbjay uses both the Tabulator and Semantic Radar Firefox plugins to do magic with RDF in his browser.

lcsh.info is gone, but there’s a lot to learn from this paper. I wish I’d seen that earlier.

Re-imagined a bit, anyway. Why browse a vertical list of results when you can flip through them like pages in a book (or album covers in iTunes). Searchme on the iPhone and iPod touch does just that.

As you type your search term, icons representing rough categories appear, allowing you to target your search and helping people who’re searching for information about pythons the snake avoid results about the programming language. Though, in practice, the category results for “python” include “computer programming”, “web development,” but no obvious category for animals or zoological queries.

The results are displayed in a cover-flow-like interface, with each website represented by a screenshot image. The result is that you can browse through a half-dozen results much faster than you could individually load each page. Unfortunately, the search results are poor compared to Google or Yahoo (to engines I use regularly), and you’ll likely find yourself having to browse a few sites before you find your answer.

Despite these annoyances, and a few bugs, Searchme is a good reminder that new technologies offer new affordances. Cover flow offers a very simple method of exploring a dataset. I’m not convinced that web search results are well suited to it, but I can imagine a few datasets — photo archives, for example — that might be.

wp-edu, the WordPress for education mail list has launched. Join up, catch up on the archives, and set it up at your school.

Declaration of Metadata Independance:

We hold these truths to be self-evident, that Metadata is essential to all Users, and that the Creation of Metadata endows certain inalienable Rights, that among these are the right to collect, the right to share and the pursuit of Happiness through the reuse of the Metadata… (read more)

Via.

I’ve wanted a Wikipedia API for a while. Now I might’ve stumbled into one: commons.wikimedia.org/w/api.php. It doesn’t do exactly what I want, but it might yet be useful.

Is xFruits a worthy replacement for Yahoo! Pipes?

The Chronicle’s Tech Therapy podcast last week featured Libraries vs. IT Departments. (Via.)

When Phoebe Ayers isn’t hanging out at ROFLcon she’s probably doing something related to Wikipedia, so I’m looking forward to reading How Wikipedia Works: And How You Can Be a Part of It.

Extra points: Phoebe and her co-authors somehow convinced their publisher to release the entire work under the GFDL, the same license Wikipedia uses. You could read the entire thing online for free, but that’s the easy part. What will you do to return the value? (Remember, Andrew Keen is watching.)

From Justin: real-time flight tracking. You can even overlay it on Google Earth. None of them as pretty as Aaron Koblin’s Flightplan, though.