I've been reading Alvin Reyes' recent blog post about "alt" text for images, and how you should manage it in Tridion. That triggered me to think about a few of these issues, and I'd like to respond to Alvin's post, and perhaps also broaden out the discussion a bit.

The typical scenario he describes is that you have a "content" component which links to a multimedia component representing an image. In the web site output, of course, the alt text needs to be associated with the image, but Alvin raises some questions over where we should manage this data in the CMS. He starts off by saying that the "Common practice" (of putting the alt text in the metadata of the multimedia component) is wrong, and that instead you should put the alt text in the content component. OK - so he then goes on to show that he understands the practical benefits of keeping the alt text with the image, and suggests a couple of other approaches, including putting the data in both places.

What Alvin describes may indeed be good practice for some organisations, but so much depends on the requirements. I think perhaps the most important take away from this discussion might be that unless you can elicit accurate requirements, you are going to get this wrong. Let's face it, the "common practice" technique is pretty solid. You have metadata that describes an image, so you manage it with the image. This isn't just an 80:20 rule, it's a 99% rule. The cases where you need multiple descriptions of the same image are rare in practice. Perhaps the obvious exception is where you're using the image as a link. Stop there a second... I mean the case where you're using a meaningful image to link to something and the description of your linking image is inappropriate for your link target.

OK - before I get carried away, let's deal with this. Firstly - there ought to be various distinct multimedia schemas in your site, intended for different purposes. Having a single "Image" schema is probably an anti-pattern. Tridion schemas have great support for constraining the choice of multimedia schema in helpful ways. (Although maybe it's weaker in Rich Text format areas). If you really, truly have a site where simple 80:20-rule alt text down on the bare metal doesn't work, then you can probably isolate the difficult cases to a particular image type or types. Obviously, you will want a schema for layout images (Lime green bullet, gothic top left corner, and so forth.) This schema should not allow for alt text, and the templating should always emit an empty alt attribute.

Then let's imagine we have a News schema, which allows for a relevant picture. When displaying the detail view of the news item, you obviously want the alt text that describes the image. "President Obama concedes defeat to the President of the NRA, at their meeting at the Whitehouse". When you render the thumbnail variant of the same image in a list of news items, the alt text will tell you that it's a link, and use the title of the news item. "Link to: Hope dies for gun control as Obama confirms his commitment to 2nd Amendment". You don't need to store any extra metadata to do this. Your multimedia schema (News event image) has exactly the metadata it needs, and your news item schema allows you to use multimedia components of this schema for adding a picture of the news event being described. (Should you also have a distinct schema for pictures whose only purpose is to look nice?)

Of course, there are sites where a high value is placed on accessibility, and where they'll do everything. Often these are government sites, or those of large enterprises that can afford the investment in keeping a squeaky clean image on how they treat the disabled. This still isn't the majority, though - by any means. The rest don't want to pay for it, and by that, I don't just mean they don't want to pay for the extra complexity in implementing all the bells and whistles. Perhaps even more to the point, they don't want to invest the extra effort that it will cost their "content" department to actually work with all this extra detail. To achieve an accessible web site, you need more than just clued-in technologists. The web content people need a pretty sophisticated understanding of the issues too. For alt text, they need to know when it should be empty, and they need to know how to write it with the right amount of context when it is needed. Generally, if you have optional fields in your schema, you'll need fairly good editorial control to ensure that it gets filled in at all. When your implementation has some sort of fallback mechanism that puts in a default when you leave it empty, how are you going to deal with that? An extra checkbox next to the alt text field that says "Yes, I really meant to leave it empty". (I'm joking, but if you're doing the site for a blind peoples' charity, you might even go to these extremes). So even with all the complexity built in to your implementation, you're still going to have to send all the content workers on courses where they get to play with screen readers and live for a day like the visually disadvantaged part of their audience, and then you're going to have to repeat this often enough to keep it real next year, and the year after.

So - most organisations aim way lower. If I had a Euro for every professional web content worker I've met who didn't know the difference between title text and alt text, I could probably go on a nice holiday. This is by no means a tirade against the content workers. Most of them are working very, very hard just keeping their sites updated, and it's no wonder they just want the CMS implementation to magically take care of all that accessibility stuff. If you put a separate field for alt text and title text, or a separate field for alt text in a different context, they will probably write something straightforward, paste it in all those places, and then curse you (again) for an idiot for making them do that last step (again). And then they'll go home and tweet that they hate Tridion.

So again, it's about the requirements. Requirements, requirements, requirements!. If your organisation is very, very keen on getting this stuff right, then your shiny new implementation isn't going to help them unless they plan for those training sessions for the web team, and then allow the content team to adjust their estimates on every new piece of content to allow for all that extra effort.... It's half past four in the afternoon, and someone in marketing or communications has emailed them a word document with the text of the press release that absolutely must hit the site at midnight tonight. Knocking-off time is half past five, but what the heck, we'll stay late to show commitment, but really, if I'm staying late it had better be for something more profound than faffing around with the third variation of the alt text. The chief editor might just bother to look for a tooltip on the image if they are on the ball, and the rest of the team are already in the pub. Whatever!

So really - if you are in that elite group that wants to do full-on accessibility, and understands and is willing to deal with the implications, good for you. Alvin's "ultimate practice" might be for you, but that'll be OK, because the team will know enough to be able to work with it. If, on the other hand, you want to do the right thing (so presumably you'll go at least as far as some moderately good alt text) the best bet might be to just stick with putting the alt text in the image component, and re-using this text for the title attribute. At least, make sure you have the requirements discussion, because the costs are very clearly not just one-off implementation costs. Even that copy-paste-paste-paste cycle will cost them time and patience every single day. Anyway - my point is that the web content management implementation needs to support the actual practices that people want or expect to follow in their day to day work.

While we're on the subject... when exactly was the last time you sat down with the person responsible for the web site and discussed the need for quote tags, or longdesc attributes? Isn't it about time?