This week I had the privilege of being given an expert guided tour through a technical documentation site running on MediaWiki. It’s a very cool site. These are the notes I wrote up after the session. I hope you find them useful! It was very interesting for me to have such a detailed look at documentation running on a wiki other than Confluence.
Rachel Perkins walked us through the Splunk product documentation, running on a customised MediaWiki platform. She and the rest of the Splunk documentation team have designed customisations to support their requirements. As well as the tools visible to the general reader on the documentation site, Rachel has added tools that make life easier for the authors on the wiki. In particular, authors can define the product manuals, tables of contents and product versions in text files, with metadata associated with each manual and version. Writers can also branch and inherit documents when a change is required for a specific product release.
Here are a few of the things that struck me about the site, even before Rachel walked us through it:
- splunk>docs: What a cool name!
- A very solid, comprehensive documentation site.
- Left-hand navigation bar showing table of contents.
- Feedback form at the bottom of each page, asking “Was this documentation topic useful?“
- Navigation at the bottom of each page, pointing to the start of a section and the next topic.
- PDF export option available on every page.
- Tool at the top allowing you to select the product version and manual that you want.
- Some user comments on the pages, but not very many. (Rachel later explained that they would like to encourage more comments. She wonders if the blue line at the bottom of the page discourages readers from scrolling further down.)
- A separate wiki for general discussion and community input.
- A section called “Hot Wiki Topics” at the top left of the wiki pages.
Rachel explained that the primary goal in selecting MediaWiki as their platform was to make the documentation as collaborative as possible. This platform made that easy. It is also important that there is a lively community of developers.
The authors value the “live” aspect of the wiki. The content is immediately available, as soon as you save it. At first, authors were a little disconcerted that there was no lengthy publication process. Rachel estimated that it took about two months for them to get over their apprehension. The advantages are that readers always have the latest information, and writers can fix problems at any time. Rachel monitors all updates via RSS feeds, which are a standard feature of MediaWiki.
- PDF exports. All manuals are available as PDF files. Customers can generate them on the fly, for the entire manual. The way it works is that the tool caches the PDF file and serves it up to each reader. It only regenerates the PDF file if there has been a change to the content. This functionality is provided by an open source plugin that Splunk has customised. PDF versions of the manuals are very popular.
- Documentation structure. Splunk have added wiki customisations to support the division of documentation into manuals, chapters and pages. To create the structure, writers edit a text file. One text file defines the list of manuals shown in the dropdown list at the top of the screen. Another text file defines the table of contents of each manual. If you load that page in view mode, you can click through from the table of contents to each page. If the page does not exist, then the tool creates a new shell page and populates it with template content.
- Versions. Similarly, there is a text file to define the versions of the product. Each version has metadata that defines its status, such as
- Branching and inheritance. What is branching? Let’s assume you need to change a page because the functionality it describes has changed in the latest release. You would then branch the page, so that you leave one branch documenting the earlier version of the product and another branch for the new version. One page can apply to many versions. You can branch a page, a set of pages, a manual or the entire documentation suite. Splunk have created a tool called the “Branch and Inheritance Console”, where you choose the version you are branching from and to, choose the manual or pages that you want to branch, and then complete a number of selection fields to create the branch. There is no awareness of the content itself. So, for example, if you change the content of an earlier version, you may need to apply the same change to the later versions too, and you would have to do that manually. Note that a page may or may not need changing for a particular version. Note from Sarah: This is very neat! Permission control (via MediaWiki’s access groups) is by version. The general public will only see the newest version when you change its status to
released. Everyone can see the versions numbers across the top of the page, indicating which version(s) the page applies to. This means that all contributors know which version to edit.
- Context-sensitive online help. This is another wiki customisation, done via tags (labels) on the pages. Each page can have any number of tags, and those tags form a kind of unique help ID for the page. Rachel recommends a minimum of three tags to ensure a unique combination. There is a file that maps the tags and product version on the page to the link from the application screen. This means that writers can move and rename documentation pages without changing the help links in the product. If a help link points to a non-existent mapping, it goes to a default page.
- Search. The search backend is a wiki customisation. They use the Google Mini Search Appliance. Rachel wants to improve aspects of the search, for example it does not recognise word stems. It’s pretty neat, though. It returns results for the version that you are currently in, or for the latest version if none is specified. Splunk uses the same backend for all its sites, not just the documentation. This means that you see search results from various sites, such as the documentation and the wiki, all on one search results page. A very nice feature is the “Search tips” panel that appears at the top of the search results sometimes. Rachel’s team have written a paragraph of tips that matches certain search terms, and will appear in the panel if someone searches for a matching term. People really like this, Rachel says.
- Feedback form. There is a feedback form at the bottom of each page, asking “Was this documentation topic useful?” The resulting feedback is mailed directly to the documentation team. Each team member responds if they have something useful to say. They love getting this feedback. The tool does not use any external service to collect the feedback. Rachel is not sure whether it is part of MediaWiki or a plugin.
- Left-hand navigation bar. This is a customisation, not a standard part of MediaWiki.
- New version of the platform. Rachel and the team are working on a 2.0 version of the platform. It will support multiple products (currently it only supports one) and will have a new look and feel.
Rachel mentioned that she is very happy with the documentation platform. I am very impressed too. 🙂
Splunk is in the process of making the customisations (all except the search) available as an open source project, headed up by Taylor Dondich. The open source project, not yet publicly available, has a working name of “Ponydocs“. Cute!
Thank you so much for a great walkthrough, Rachel! And congratulations to the Splunk team for this excellent set of documentation.
This week I’m attending the WritersUA 2011 Conference for Software User Assistance in Long Beach, California. One of the sessions that I attended today was an Adobe lab, run by Scott DeLoach, Kevin Siegel, Laurie Edelman and others.
During the conference, there were a number of “lab” sessions. The labs offer a number of computer workstations and software that we could use for hands-on experimentation. A few experts wandered around offering help and advice.
My aim in attending the Adobe lab
I decided to take a look at the latest version of RoboHelp. I have used RoboHelp extensively in the past, but not since RoboHelp 8.
In particular, I wanted to explore the new features in RoboHelp that allow you to gather feedback from users, and allow users to share comments among themselves. I had heard and seen this demonstrated in various places. The idea is that you have an online help system, shared amongst your users, even installed individually on their machines. Each user can add comments to their own version of the help, and share those comments with other users via a central server.
Exploring RoboHelp’s comments feature
In the lab there was no RoboHelp Server available, so we could not see the server side of the functionality in action. But I did manage to explore how it would work. Here’s what I did.
In RoboHelp, create your online help topics as usual. For the lab session, I used one of the sample projects supplied by Adobe.
The next step is to generate your help output, selecting Adobe Air as your output format. The Adobe AIR output is the only format that supports comments submitted by users. To generate the AIR output:
- Choose one of the a “single source layouts” available. A single source layout basically corresponds to an output format. For this lab session, a few Adobe Air layouts were available. I chose “Adobe AIR – Role Based”.
- Double-click the layout in the “single source layout” window. You get a popup where you can configure various options for the chosen layout.
- On the “General” tab, choose an “Output Type” of “Adobe AIR Application“. This is the only output type that will allow comments from users. The other output types are hybrids that weave a Flash solution into an HTML web-based help system.
- You will need a digital certificate. If you don’t have one (which is likely at this point!) you can generate a self-signed certificate for use while testing. Eventually you will need to get a certificate from a real certificate authority (CA). To generate a digital certificate, click “Create” next to “Digital Certificate”.
- Supply a publisher name of some sort.
- Specify a password for the certificate.
- Next to “Save As”, specify a path and name something like “C:\cert.p12”.
- Click “OK” to generate the certificate.
- Click the “Collaboration” tab.
- Select “Enable Topic Rating”, “Enable Commenting” and “Enable Comment Moderation”.
- Set a password that the comment moderator will use.
- At this stage, it would be good to specify the RoboHelp server to enable server-based storage and sharing of comments. I could not do this during the lab, because there was no RoboHelp server available.
- Click “Save and Generate” to create your online help system.
- Click “View Result” to see the resulting help files.
The AIR application will start up, showing your help topics in the help system. Now you want to add a comment to a topic.
- Click the comment icon near the top left of the screen. (The icon looks like a couple of comment bubbles.)
- A comment area will appear at the bottom of the topic.
- To add a comment, click the plus sign.
Adding a comment:
Comments are displayed in reverse chronological order:
This looks pretty cool. As far as I could see, there’s no opportunity to add graphics or formatting to the comments. They’re plain text. Also, they’re not threaded – you can’t reply to a specific comment. Still, once a server is hooked up it is pretty cool for people to be able to share comments from their desktop apps. I’d be interested to see it in action, and to hear any experiences other technical writers may have had with this feature.
This week I’m attending the WritersUA 2011 Conference for Software User Assistance in Long Beach, California. These are my notes from the session called “Hotrod My Help” presented by Leah Guren. If you find any inaccuracies, they’ll be mine.
Leah’s presentation style is both informal and professional. She started out by introducing herself as “this deranged woman”. She then became more serious quite quickly. Her talk was about design of online help systems. She used the metaphor of hot rodding – talking about the paint rather than what’s underneath.
Introducing the topic
Leah’s talk was entirely tool agnostic, and described techniques that we can apply to our documents with very little effort.
Nevertheless, Leah pointed out, this stuff matters for a number of reasons:
- Design should match the subject matter.
- Better design improves usability.
- We’re not talking about making things look pretty. We want to make sure that what we develop will communicate as effectively to users as the words we choose.
- When people see help with a sloppy design, they’re less likely to trust the information.
- Design should support the meaning of the content.
- It should make it easier for people to find what they’re looking for, to recognise it and to use it.
Leah also pointed out that if your readers are noticing the design, then you have made a mistake. (As a side note, I tweeted this point and received an interesting reply from Bill Kerschbaum.)
The design concepts
Leah introduced us to some design concepts, the jargon. Mastering the terminology lets you defend you choices with more authority.
Leah introduced us to the acronym PARCH:
- Proximity – When elements are close together, we recognise that they are related.
- Alignment – Choose where you position elements on the screen, to develop a meaningful design.
- Repetition – Repeated visual patterns help people to access and remember the information efficiently.
- Hierarchies and dependencies
One stylistic mistake that violates this principle is the “floating heading”. That’s when you have a heading with the same amount of space above it as below it. Instead, there should be less space below the heading, to bring it closer with the content it describes.
Use alignment for a less sloppy design.
Consider things like colours, icons, placement. Stylistic repetition reinforces the information for the user and gives them more confidence.
We need contrast so that people can see the differences in meaning between different elements on the page. For example, links should look different from other text. So should headings, and layered information. The rule of contrast is that it must be meaningful. Readers don’t care how you do it, whether you use a list or headings or bold text. Just be consistent and make it meaningful.
This defines how elements of information on the page are related to other elements on the page. Consider the use of white space and indentation.
Other concepts and techniques
- White space – Consider margins, space between lines, hanging indents, paragraphs etc. White space is critical. If you don’t have enough white space, the text is not readable. Users flee from densely-packed screens.
- Plumb lines – Draw vertical lines at every point where an information element starts, to make a clean design. Good clean design steps into the background. Bad design is noisy.
- Indents and text wrap – Indents help the concept of contrast. Good examples are bulleted lists.
- Paragraphs – Leah recommends the following design techniques for paragraphs, to enhance readability:
- Flush left, ragged right.
- Single line spacing.
- White space between paragraphs.
- Chunking – Use white space intelligently, to show the logical visual chunks of information.
- Nesting – Nesting relates to hiearchies and logically related information. Leah mentioned “illegal” nesting. The rule in technical communication (a tacit contract with the user) is this: if I’m going to break the information down into logical categories, there must be more than one category. For example, you can’t have just one item in a bulleted list, or just one level-2 heading.
Examples – applying the concepts
After introducing all the theory, Leah moved on to examples of web page makeovers, using her hotrod metaphor. She took us through some real documentation and web pages from various sites, and applied her principles to each one.
These are some of the mistakes she fixed up:
- Messy text.
- Dashboard that was not designed to be meaningful.
- Screenshots taken without care and without tidying up afterwards.
- Sloppy text wrap, that probably happened when pasting content from elsewhere.
- Inconsistency in capitalisation, punctuation, and so on.
- Long lines that force the user to do horizontal scrolling. A typical culprit here is information in tables. This is easy to fix, but easy to overlook.
- No visual cue that content continues below the fold. Readers may think they have reached the end of the steps, without realising they need to scroll down. One way to fix this is to include a “Start” and “End” icon.
- Help window opening on top of the application by default. The solution is to make the default opening position at top right, for example.
- Concordance (repetition of same words) in the table of contents (toc). This is very distracting. If the toc entries are very long, people will not see the meaningful bit in the toc window.
- Use of blue (usually used to denote hyperlinks) as an emphasis or point colour.
- Too many layers of hierarchy (too much nesting) in a table of contents.
- Links in the text. Don’t put links in the text unless they’re popups or dynamic HTML. Keep all links off to the side or at the bottom. People get distracted by links, or click them and get lost.
- Empty topics.
Help authoring tools
If the problem we’re addressing lies within a tool that we use to develop help systems, then the HAT and DITA tool developers need to fix it. We need to be involved in the design of such tools.
Throughout the presentation, Leah had “bonus rounds” where she asked us to name concepts that the pages were violating. She promised us “valuable prizes” (said with a grin) which we could collect afterwards. This was great for getting audience participation going.
Thank you for a lively session, Leah!