Posts from Wednesday, July 21st, 2004

Thomas Jogin recently posted an interesting entry that asks about HTML headings.  He starts out this way:

For some time now, something has been bothering me about the common way of marking up a document structure. First of all, it doesn’t seem like there is a consensus on wether or not to use multiple H1 tags, or any at all.

I agree: there isn’t a consensus on that point.  There is also a lack of consensus on his second question:

Secondly, since H1-H6 are supposed to denote the structure of a document, do they really have a place in the side column as section headlines?

In fact, there’s very little consensus on heading use in general, except that you should use them instead of font size="+3" and that sort of thing.  Headings make for more visibility in Google than does presentational markup, after all.

In considering Thomas’ question, Andy Budd goes to the specification, which is always a good place to start.  As he points out, HTML 4.01, section 7.5.5 states:

A heading element briefly describes the topic of the section it introduces. Heading information may be used by user agents, for example, to construct a table of contents for a document automatically.

There are six levels of headings in HTML with H1 as the most important and H6 as the least. Visual browsers usually render more important headings in larger fonts than less important ones.

So as far as a semantic definition of the heading elements goes, all we have is that heading levels indicate degrees of importance.  Nothing about what order they have to be in, or whether you can skip levels, or anything else besides the creation of a spectrum of importance, as it were.  Because that’s all we have, there is plenty of room for people to fill in their own interpretation of what’s best.  I approach headings as merely indicating a level of importance, and don’t bind myself to a decreasing numeric order.  That’s my take on it; others feel differently.  Let’s take a look at two cases where my approach led to very different results.

The first case is Netscape DevEdge.  Here’s the markup skeleton for the first part of the home page as I set it up (the actual markup as of your reading this may be different):

<h1>
 <a href="/" target="_top"><span>Netscape</span> DevEdge</a>
</h1>

<form action="/search/app/" id="srch" method="get">
<h4><label for="search-input">Search</label></h4>
(...inputs...)
</form>

<h2>Netscape 7.1 is now available</h2>
(...paragraph text...)
<h2>The DevEdge RSS-News Ticker Toolbar</h2>  
(...paragraph text...)
<h3>Recent News</h3>
(...list of links...)

Here, the order is h1-h4-h2-h2-h3.  I guarantee you that some readers are feeling a deep outrage right now, because we got more than one piece of feedback berating us for putting the headings ‘out of order’.  The h4 was clearly out of place, we were told, and needed to be fixed.

The placement of the search markup in the document was dictated by our preferred document linearization, however.  In an unstyled view (a really old browser, say, or a cellphone text browser) we wanted the masthead to be immediately followed by the search feature, and that to be followed by the main content.  The main content was in turn followed by the sidebars, which was followed by the navigation and configuration links, and finally the footer.  So we weren’t about to move the search markup just to make its heading level fit into a descending-number hierarchy.

We could have elevated the heading to an h2, but that struck me as being a poor choice.  Is the “Search” title really as important than the top headlines on the site?  I certainly didn’t think so.  My only other option was to change the h4 to some non-heading element and style it that same as I had the h4  That’s certainly easy to do, since with CSS you can make any non-replaced element look like any other element, but that would mean the search area had no heading at all.  That also struck me as a poor choice.

So I took the route I did.  I suppose I could have, as Richard Rutter suggests, put in some extra headings to increase the structure and then “switched off” their display with CSS.  That would mean inserting an extra h2 and h3 between the masthead and the search.  What would they say?  Personally, I’m not thrilled with that idea either, at least for the DevEdge case.

The second case I’d like to consider is meyerweb.com itself.  Here’s the basic structure of the home page, at least as of this writing:

<div id="sitemast">
<h1><a href="/"><span>meyerweb</span>.com</a></h1>
</div>
<div id="thoughts">
<h3 id="thtitle"><span>Thoughts From Eric</span></h3>
<div class="entry">
<h4 class="title"><a href="..." title="...">entry title</a></h4>
<h5 class="meta"><b>entry date</b></h5>
</div>
<div class="entry">
<h4 class="title"><a href="..." title="...">entry title</a></h4>
<h5 class="meta"><b>entry date</b></h5>
</div>
</div>

Here, I skipped the h2, but other than that everything’s in a descending-order hierarchy.  In this case, such an approach was a fairly natural fit to the information being presented.  I don’t have an h2 on the home page because it’s reserved for page titles, such as the “Eric A. Meyer” at the top of my personal page.  The page title of the home page would be “Home Page” or “Main Page” or something equally silly, so I left it off.  (And yes, there are b elements in my markup, and I’m doing that on purpose.  I’ll talk about it some other time.)

Meanwhile, in the sidebar, the section headings (such as “Distractions”) are enclosed in h4 elements.  That makes them as important as entry headings, and I’m okay with that.  I might have chosen h5 instead, but probably not h3.  The one exception is “Blog Watch”, which is an h5 element.  I don’t have any rigorous logical explanation for this; it’s all based on gut feelings.

In the end, I don’t think there will ever be a heading consensus.  XHTML 2.0 attempts to bridge the gap by adding the generic h (heading) element, which helps create a descending-order hierarchy when combined with nested section elements.  What bothers me is that the Working Group hasn’t decided whether or not h1 through h6 should be deprecated; to do so would in effect attempt to force a consensus in favor of a descending-order hierarchy.  That would mean that headings were no longer as much about importance as they were about having a document that’s structured like a specification, or maybe an e-book.

In the here and now, though, we’re left to ponder these questions and arrive at our own conclusions.  Each answer will depend on the temperament of the ponderer, and the project upon which the pondering is predicated.

• Pick A Heading was published on Wednesday, July 21st, 2004.
• It was assigned to the Standards category.
• There have been sixty-five replies.

A few people have written me over the past week to point out that I could have saved myself a lot of time when I transferred files over to my new Mac laptop.  Standards and accessibility warrior DeWayne Purdy was one of the first, and wrote:

…there’s a better way than ethernet to transfer  files between two Macs, using Target Disk Mode. Connect the two Macs  together with a firewire cable, with one Mac on and the second one off.  Then turn the second Mac on and hold down on the T key. A firewire  symbol will come up on the screen of the second Mac, and the HD name  will show up on the first Mac, just like it’s an external hard drive.  The files can then be transferred at firewire speeds, much faster than  via ethernet.  (These are condensed instructions, I’d look up the full  instructions before doing it… there’s a few little, but important,  details, like making sure you drag the icon of the second Mac to the  trash before disconnecting or shutting it down).

So I looked up the full instructions, and here they are.  They’re not much longer than DeWayne’s summary, actually.

• FireWire Transfer was published on Wednesday, July 21st, 2004.
• It was assigned to the Mac category.
• There has been one reply.

It seems I was a little unclear in my recent post about floats, and compounded the problem with the post title.  A number of responses said, basically, “maybe they’re bad for layout but what else is available?”  I didn’t mean to suggest that nobody should use floats for layout, or that using them in that manner is somehow wrong; my apologies if I did so.  There aren’t too many alternatives; in fact, there are precisely three general standards-based layout options.

1. Floats
2. Positioning
3. Tables

Each of these options has its benefits and drawbacks.  All I was trying to say is that Andrei was dead on in his feeling that layout floats feel like something that have been bent to a purpose for which they weren’t intended.  That’s true.  Floats are simple, and were meant to do a very limited thing.  Push them hard enough into other roles, and they become a touch balky.  If nothing else, their source-order dependencies make them less flexible than one might like, and variances in browser handling of floats (many of those variances having to do with how one should deal with floats in their intended role) just makes the situation more frustrating.

One could argue that all three options are non-layout technologies that have been bent to layout purposes, but positioning at least was intended to be a layout mechanism from its inception.  It suffers from enough glaring flaws and browser bugs that it can’t be considered any better an option than the others, though.

In a move that may help matters, Shaun Inman (he of the Inman Flash Replacement technique) has just published the article Clearing Absolutes Again, which  further develops his attempts to give positioned elements a clear-like capability.  This is a very welcome development.  The IPC (Inman Position Clearing) technique may not be perfect— I haven’t had time to dig into it, so I don’t know— but it’s a fifty-foot neon arrow pointing in the direction things need to go.  Since we can’t expect this sort of capability to be added to CSS in the near future, let alone have enough browsers support it in the next few years, scripting our own fixes is the only reasonable solution I can see.  For another great big neon arrow of this kind, see Dean Edwards’ IE7.

And yes, I said “tables” back there.  If you can avoid using tables for layout, then by all means do, but it isn’t always possible.  In such cases, as long as the tables use as little markup as possible, and the markup validates, then I don’t see the problem.  Never have, really.  My whole objection to tables-for-layout has been, and continues to be, that most table-based designs use way too much markup.  Often you see markup-to-content ratios in the vicinity of 3:1.  That’s just icky.  I much prefer to see the inverse of that ratio, or at least something below 1:1.  CSS helps immensely in getting there.

Update: see Shaun’s new new post on the topic, “Absolute Clearance“, for another step in the evolution of his technique.  Ever forward!

• Floating Points was published on Wednesday, July 21st, 2004.
• It was assigned to the CSS category.
• There have been eight replies.