How We Got Into This Mess

When I was hired at Pop Art in early 2006, there was no official Pop Art blog. The design team had their own site, which was sporadically updated and not branded as a Pop Art site. Several members of the dev team also had work blogs, but again, they were not branded. As somewhat of a blogging zealot, I pushed hard to standardize these onto one site, and in August, I got the go-ahead to set something up.

When I started looking into solutions, I was working with a few assumptions. First, that the existing blogs would stay up, and the central blog would syndicate their posts by pulling in their RSS feeds. Secondly, despite my strong preference for Wordpress, that it was not a viable option for us. That’s because we’re a Microsoft shop, so we don’t have a lot of PHP experience in-house, and we don’t have any Apache boxes to host the site on. Also, since I was assuming we wanted multiple blogs syndicated at one domain (rather than a single blog with multiple authors), there wasn’t a good way to do that with Wordpress at the time (Wordpress MU hadn’t really taken off at the time).

Given those assumptions, we ended up installing Community Server. I felt good about this because it was the product used to host all the blogs at Microsoft, and appeared to be well-supported and active.

Increasing Difficulties with Community Server

However, installation and customization were a nightmare. After a lengthy struggle to get everything set up correctly, it became increasingly clear that some of our assumptions were inaccurate.

Although we started with independent blogs being syndicated, our CEO eventually wrote a company blogging policy disallowing this, so we ended up with a Community Server site that had 20+ single-author blogs being aggregated onto a single homepage to give the appearance of a standard multiple-author blog. We weren’t taking advantage of anything the multiple-blog scenario was originally set up to do.

Over time, we found out that Community Server was not only difficult to customize, it was also poorly supported. After we purchased and installed CS2007, they released two further versions, and stopped supporting ours. (When we contacted them about purchasing an updated for our license to add more blogs, we were told our version was no longer supported, and they offered to let us upgrade at a discounted price of $900. We didn’t purchase that upgrade, and three months later, they announced that they were changing their license structure again, and we would now have to pay full price for a brand new license if we wanted to upgrade.) I know that companies make money by releasing new versions, but it seems crazy that less than two years after we bought their product, they wouldn’t support us at all without a full upgrade.

To cut a long story short, the straw that finally broke the camel’s back was when we decided to integrate Twitter into our blogs. We were able to add the javascript twitter script to the sidebar to display the Pop Art feed easily, but in order to add individual twitter feeds to the profile page, we were going to need to bring in the .NET programmers, who are always really busy — and I realized that if we were using Wordpress, I could do everything myself.

There were other factors, including our frustration with the licensing situation, and the fact that we didn’t need the overly complicated multiple blog situation anymore, but the customization difficulties were the kicker. Once I was able to assure our CEO that switching to Wordpress wouldn’t lead to a repeat of this situation in a few years, we had the green light to make the switch.

Exporting Posts from Community Server

Getting our posts from Community Server to Wordpress was our first hurdle. Out of the box, there’s no way to make it work, other than having Wordpress scrape the RSS feed — and that would lose all the comments. Thankfully, other people had already done the legwork for me. Bear with me here.

Aaron Lerch wrote a great post in August 2007 about switching from Blogger to Wordpress using BlogML. I already knew from when we originally set it up that Community Server supports BlogML, and Aaron wrote a module for Wordpress to import from BlogML format.

Then, in October 2007, Rob Walling wrote a post about switching from Community Server 2007 to Wordpress. He updated Aaron’s Wordpress importer because Wordpress 2.3 switched from categories to tags.

Finally, in October 2008, Kavinda Munasinghe wrote a post about switching from DasBlog to Wordpress, and once again, updated Aaron’s Wordpress importer.

So, I grabbed the BlogML Export module for Community Server 2007, and Kavinda’s updated BlogML Importer for Wordpress, and set about my business.

Importing Posts to Wordpress

The first problem I ran into is that I couldn’t get one big export file of every blog in our Community Server installation, I had to create an export file for each one.

The second problem I ran into is that my exports didn’t import correctly. Now, I have no idea if the problem was from one of the versions of the importer, or if the exporter wasn’t working correctly, or even if the latest version of Wordpress just has a slightly different database. What it boiled down to is that I had to manually reformat each of my 32 export files.

The biggest issue was with categories/tags. First of all, all my tags were being imported as categories. I was willing to ignore this, since Wordpress has a “convert categories to tags” feature, but it was still annoying. In addition to that, however, all the categories were importing as numbers. Turns out that Wordpress was grabbing the category ID instead of the category name.

I looked at both the importer and the exporter, but I didn’t have the programming know-how to fix it, so ultimately, I just wrote a simple regular expression to reformat all the categories so that their names were in the ID attribute so they would import correctly.

In addition to that, author names weren’t importing correctly, draft status posts were being published, and a couple of older blogs where the authors had copy-and-pasted from Word were bringing in a bunch of crazy custom microsoft markup. All in all, I probably spent about 3-4 hours manually cleaning up all my export files. This all sounds like a huge pain (and it was!), but I’m happy to say that once I was done, everything imported perfectly.

Redirecting Old Posts

The final thing I needed to do was set up a 301 redirect for all the old posts. Aaron’s importer script gives you a file that you can use to build a list of redirects in your .htaccess file, but I was sure there was a better way. After a bit of digging, I found a forum post about using mod_rewrite to change dynamic to static URLs that helped me write the following regular expression to redirect from our old permalinks to our new ones.

RewriteRule
     ^([^/]+)/archive/([^/]+)/([^/]+)/([^/]+)/([^/]+).aspx$
     http://blogs.popart.com/$2/$3/$5 [R=301,L]

Note that normally, that would all appear on one line. I’ve broken it up for readability. That looks for urls like /scott-vandehey/archive/2006/11/09/post-title.aspx and redirects them to /2006/11/post-title/. With that one rule in our .htaccess file, all our old blogs posts would automatically get a 301 redirect to the new URLs.

Now that we’ve made the switch, I’m really happy we spent the time on it. We took the opportunity to redo the design to better match our main site, and the template engine in Wordpress is a real joy to work with — and well-documented, more importantly! I’m confident that we’ll be able to handle the evolving needs of our company in the future, and informally, I can say that everyone seems happier using the new system. The transition was a pain, but the results were worth it.

If you’re reading this in your feed reader, then our RSS feed has automatically redirected, and you shouldn’t have any problems. But in case you’re a stickler for official URLs, or you haven’t subscribed, our new feed can be found at blogs.popart.com/feed/atom/.

In the next few days, there are some more posts coming, including one from Mark about the new design, one about how we integrated Twitter, one about some nifty Wordpress tricks we learned, and a long one about our experience converting from Community Server 2007 to Wordpress (long story short: it makes scaling Everest look like walking up the stairs).

So please: snoop, dig, stop and smell the roses. And welcome to the redesigned and improved Pop Art Blog, it’s where the magic happens (or something like that).

Before I get into how it went wrong, let me point out what went right. First of all, we did eventually succeed in getting nearly everything we wanted done, and I’m incredibly proud of the final results. Secondly, although I experienced many frustrations along the way, the official support forum was very helpful, and I always got a response to my questions within 24 hours. If you’re going to try skinning a CS2007 site, I highly recommend getting an account on the forums, and working with a programmer, if you can (more on that later).

Finally, although I did have difficulties, CS2007 is built to support a corporate blogging environment from the ground up. Wordpress and Movable Type are better programs in terms of the overall blogging experience, but their support for multiple-user scenarios are limited at best. Wordpress MU comes close, but I don’t like being removed from the primary development branch. At the end of the day, CS2007 is clearly the best solution for corporate blogging the way we want to do it.

With that out of the way, let’s talk about why skinning it was such a frustrating experience. It all boils down to one simple problem: A complete and total lack of documentation for non-programmers. Allow me to explain.

There are basically three classes of users for a product like CS2007 or Wordpress. You have your basic users who don’t know anything about the code, and only interact with the program through the browser interface. These users have minimal documentation needs. Give them some instructions on how to make a new blog post or add their logo to the default template, and they’re good to go. Both Wordpress and CS2007 have this basic documentation.

Next, you have your programmers. These are power users who are happy to muck around in the source code if they don’t like the way something works. These users have advanced documentation needs, because they need insight into things like the database schema, program controls and common functions. Both Wordpress and CS2007 have this advanced documentation. In fact, Community Server’s comes in an MSDN-style CHM file listing all the controls that are available. My programmer cohorts assure me that this documentation is quite thorough and gave them everything they needed to know.

Lastly, you have your casual developers. These are the guys like me who are very familiar with front-end code like HTML, CSS and javascript, but don’t know any .NET or PHP. These users have moderate documentation needs. They don’t need to understand how the source code works or the structure of the database necessarily. What they need is an easy way to pull out specific bits of information.

To give you a way to compare, let me give you a simple example common to both WordPress and Community Server.

In Wordpress, to display the date a post was made, you use this code: <?php the_date(); ?> I found this out by going to the Wordpress documentation, clicking on Template Tags, and then clicking on the_date in the “Date and Time tags” section. I was given a description of the tag, a list of parameters, and some examples of how to use it.

In Community Server, to do the same thing, you use this code: <CSBlog:WeblogPostData runat="server" Property="PostDate" /> I found this out by visiting the Community Server documentation section, and searching (in vain) for information on modifying templates. Eventually, I stumbled onto the Single Value Controls page, which helpfully explains the structure that is used to get a bit of information out of the database. This page does not give a complete list of values, however. To get that, I had to backtrack using the breadcrumb trail at the top of the page to the Chameleon Control Documentation page, which links to the MSDN file I mentioned before, as well as “API Property Documentation”, which includes a link to Blog Control Property Data, which lists all the properties for the CSBlog:X control. By combining the structure I found on the Single Value Controls page and the property names I found on the Blog Control Property Data page, I was able to make a control that echoes out the post date.

Now, that’s really not a great example, because I was never trying to look up something as simple as the post date. For something like that, I would look for an example in an existing theme. I’m just using this to illustrate what a pain it is to find any information on the Community Server side if you’re not a programmer. I tried looking in that MSDN document, but it was all greek to me. Even when my programmer cohorts were assisting me and walking me through the document, I didn’t understand what I was looking at.

A more common example of the kind of thing I needed documentation for was how to display a list of weblog posts but limit it by a certain tag. I never did find any documentation explaining how to do something like that. If you need anything like that, your only recourse is the forums. For example, here is a search I did for filter blog posts by tag. You’ll notice that of the results I got, most of them have nothing to do with my search, except that they are about how to use tags. Of the posts that do seem relevant, you’ll notice that most of them date from 2005 or 2006 – meaning they refer to the previous version of Community Server, and are no longer relevant.

Everything is like this. Searching for the most simple information becomes an incredibly frustrating ordeal. The only saving grace of the forum is that it is populated by people like Ben Tiedt, who managed to answer all the questions I posted within 24 hours, and was really helpful.

So, to recap: If you want to find anything out about Community Server, be prepared to spend hours trolling the forums and documentation, sifting through the irrelevant and out-of-date results to possibly find an answer, but more likely, you’ll have to post a question yourself, and wait for a community member to help you out.

I can’t be the only one who thinks that forums should not be the primary method of sharing information. Community Server desperately needs Wordpress-style documentation aimed at casual developers (like myself) who have higher needs than the average end user, but aren’t programmers.

We launched two new Center Sites in November. Kansas City was a challenging design, with lots of absolute positioning, which caused a few problems when IE7 came out. The client was really excited about the puzzle theme. Birmingham is, I think, one of the best looking designs we've done. The diagonal navigation is really sweet.

Updates to existing sites

We helped out several clients with minor updates to their sites. In one case, we redesigned their homepage to emphasize a change in the way they're marketing to their users. In several other cases, we did some heavy-duty testing and bugfixing for IE7. In one case, we beat our heads against a display issue that the client was seeing that we couldn't reproduce. In the end, we managed to track it down to a sIFR problem caused by the client having JavaScript disabled. Or, rather, to the fact that I had forgotten to write styles for a non-JS environment. I was pretty embarrassed about the whole thing, but it served as a good reminder for the future to consider what happens in non-standard settings.

Carrier Media Center

We also launched the Media Center for Carrier. This isn't a huge section of the site, but it was a lot of fun to work on. Ben created a flash movie player, and we got a solution set up to let non-iPod listeners hear the podcasts in the browser. We've probably got more multimedia work like this in the future, which is kind a fun break from our normal work.

Project "Bloom County"

Work continued on project codename "Bloom County" – our in-house WCM solution. Andrew spent some time on a functional prototype, and I spent some time making mockups of how I think a good structured content solution would work with one of our existing sites. Then we presented the work to Management, Design, and Client Services. Their reactions spanned the gamut, and it was really interesting getting feedback based on something real, instead of the abstract concept discussions we were having several months ago. This project may or may not continue, but the discussions we had as a result of it were invaluable.

Blogging

I finally "formally" announced the blogs site to the company early in November. There was no real reaction at first, but I heard that Steve, the CEO has some plans for it in the future, so I'm excited about the potential. I applied some anti-spam measures, since I was starting to get some comment spam, but I was disappointed that we don't have room in the budget for a commercial license for Akismet, but I managed to find a good set of spam-plugins on the Community Server site. (If you can afford it, or you're running a non-profit blog, however, I can't recommend Akismet strongly enough, and now that there's a CS plugin for it, you should really check it out).

I also found a problem with DasBlog and a related problem with CommunityServer. We've got lots of blogs set up on the CS site that just function as an aggregator to get content from external blogs into the CS Portal (so that Pop Art can have a single site for all blogs). This works pulling content from the RSS feeds on my WordPress blog and from Kelly's CS blog, but it was choking on Andrew's DasBlog.

What was happening was that all the categories on Andrew's blog posts were coming through mashed together, so instead of seeing three categories, "Programming," "ASP," and "Humor," I would get one category, "Programming;ASP;Humor." This was really screwing up our tag clouds, so I switched Andrew's blog to pull from his Atom feed – only to find out that CS doesn't support aggregating from an Atom feed. This is disappointing, but it's hard to be mad at CS, since the real problem is the messed up categories from DasBlog's RSS feeds. Andrew is going to hack his feeds at some point, but until then, we've got no feed for his blog.

Book Reviews

You may have noticed me posting some book reviews recently. That's because I took advantage of my downtime in November to catch up on my reading. I had three new books to read, Bulletproof Web Design by Dan Cedarholm, CSS Mastery by Andy Budd, and Web Standards Solutions by Dan Cedarholm. They were all pretty good, but mostly it was just reassuring to see that I'm already following best practices for the most part. I've still got two books to read – DOM Scripting by Jeremy Keith, and Naked Conversations by Robert Scoble.

We launched four SelecTrucks sites this month, and all four feature landscape shots. Denver has a nice big mountain, and lots of wet-floor reflections on their photos. Cleveland has several landmarks in a photomontage, Twin Cities has a great landscape shot at sunset, while Seattle has a very nice cityscape, including the space needle. All four were fun to work on, and we're really getting our production process down pat.

LifeWorks NW

We managed to push LifeWorks' site into Beta, and then started the process of responding to customer requests. As before, this site remained challenging, but I really believe that once it launches, all the hard work will be worth it.

PICA email

We launched the new PICA.org site, and pulled together two email newsletters to promote it. I wasn't involved in the site itself, but I did produce the emails. I've pretty much become the company's go-to guy for email production, and I have to admit that I kind of like it. There's a certain perverse thrill to producing old-school table-driven font-tag-laden pages, and since these are always single-page emails, they go quickly, usually just a few hours. So it can be a nice break from my normal work.

Australasian

Trying to keep my billable hours up, my manager passed on some site edits for an older site that Pop Art did for Australasian College of Health Sciences. Like with the emails, this was a challenge to get in and make my changes while disrupting the existing code base as little as possible. These were mostly quick content edits, but there were some crazy moments slicing up graphics in Fireworks to edit the navigation.

Design/Dev team

During August, the dev team and the design team started having meetings to hammer out a better working process. Those went very well, and I think everyone walked away feeling like there were some real simple things we could do to improve communication and efficiency between our teams. After that, we kept having the meetings to discuss what our company wants from a WCM solution. We're still discussing this today, but these early meetings helped lay the groundwork for how we would be working today.

Blogs

My big side project during August was to get the Pop Art blogs site set up. After doing some research, we decided to use Community Server as our default solution (though individual employees can set up their own blog using whatever software they want). My blog posting dropped of during this time, as we researched ways to get my blogs posts out of DasBlog and into Community Server. We finally accomplished this using a script that Scott Hanselman himself provided us, which exported from DasBlog to the BlogML format, which a Community Server plugin let us import. By the end of the month, we had the new Community Server install running, and my old blog imported!

DasBlog wasn't a bad program, but coming from a background in WordPress, it was frustratingly limited and very rough around the edges. I've already detailed my frustrations with DasBlog, so I won't repeat myself here, except to say that the only solutions I've found to my problems are to either fix them myself (not practical, since I'm not a programmer) or wait for a new release (which, to be fair, is about to come out, but it's still not a good answer).

Community Server is a much more polished solution, which is to be expected, since it's the program that's running blogs.msdn.com. Where it really shines, however, is as the backbone of a blogging community. Our goal was to allow Pop Artists to sign up for an account, and have a basic blog be created automatically. Then if they want to muck around with advanced options or skinning, they can, but they also have something that just works out of the box.

For me personally, the two best features are the functional posting page (no more crappy text editor – CS has a more wordpress-like text editor which automatically wraps my paragraphs in <P> tags, and gives me a nice little HTML editor window to do things like adding that "P" tag back there) and the improved control over the skinning. DasBlog was VERY limited on what I could and could not modify in the skins, and CS removes that limitation.

Of course, it's not perfect. CS delivers way more out of the box than we actually need, and it's not easy to disable the extranious bits. We don't need forums, or photo and file galleries, and the included RSS reader is just confusing. Giving my druthers, I would just turn those off and only use the blog engine (this may actually be possible, I just haven't found it yet). It's also stupidly difficult to move the blogs into the root directory (by default, blogs are in a subdirectory, resulting in the irritating URL "http://blogs.popart.com/blogs/scottv/"), involving physically copying files (which needs to be repeated after each upgrade), and modifying SiteUrls.config. Also, it seems that I have to link to my blog as http://blogs.popart.com/scottv/default.aspx – if I removed the default.aspx, I get a 404 – which is dumb, since default.aspx is defined as the default page in IIS by default.

Finally, and most irritating, there is no import/export functionality by default (also true of DasBlog). In order to bring over my blog, we had to install the BlogML Converter Control by Keyvan Nayyeri, which adds an import/export tool for the BlogML format. This got us halfway there, but despite information to the contrary, there's no BlogML exporter for DasBlog in v1.8. We finally got the problem solved after writing to Scott Hanselman himself, who very kindly provided us with his work-in-progress BlogML exporter from v1.9. With a bit of tweaking, we managed to get all the blog posts and categories imported, more or less (none of the categories are applied to any of the posts, but by this point, we'd put a week into it, and no one cared enough to fix it, so I'm just going to re-apply the categories by hand). To be fair, this should get a lot easier, once DasBlog 1.9 is out, but right now it's a royal pain in the butt.

Still, I'm up and running, and planning to start blogging again right away, and I'm pretty happy with Community Server so far…