Re: [uf-new] Microformats for hidden data
Brian, thank you for the very detailed response. I do understand this better now. I completely agree about the issue of out of sight, out of mind. It's exactly the problem that applies to web analytics data now. What I want to do is to bring the data a little more out into the visible world. What typically happens on big enterprise sites is that they have an analytics product which requires certain per-page metadata, such as a page name and category. This is different from typical installations of the free tools like Google Analytics, partly because these are larger, more complex sites with deeper analytics needs, and partly because they often have horrible legacy URL structures which makes it impossible to just record visits to URLs. There is a lot of existing deployment of these tags, see for example data at http://www.jgc.org/blog/2009/10/some-real-data-about-javascript-tagging.html Our overriding interest is in making the data available to more than one tag on the page, so that the data doesn't have be declared multiple times in different formats. It's certainly possible to do this purely in JavaScript, where the data is currently declared. But the secondary goal is make the information a bit more accessible for the people who are responsible for the content. Many of them are editors who are reasonably comfortable reading html, but won't touch JavaScript because it is programming. Hence the interest in potentially using microformats. Right now, the editors who are responsible for populating the data, and the analysts who are the audience, commonly have no access at all to check whether it's correct or correct any issues. Would you trust a smaller set of data that has a higher probability of being accurate, or loads of hidden data that has a higher probability of being inaccurate? I agree completely completely with this sentiment. But my question is, given that there is data which is already hidden, crufty and out-of-sync, can we do something to shed a little more light on it? I hope this helps explain where we're coming from. Fiann ___ microformats-new mailing list microformats-new@microformats.org http://microformats.org/mailman/listinfo/microformats-new
Re: [uf-new] Microformats for hidden data
Brian, that's exactly what I am hoping to do, you have captured it precisely. hAtom gives a lot but not all of what I am looking for (my baseline is the fields that are common to all the major web analytics products). hAtom is focussed on blog posts rather than generic website pages, and I am not sure it is an exact fit, but the core concept is very similar. The reason I am interested in using microformats is that if by using a standard, I can turn your suggestion of var page = $('.entity-title'); into hAtom format var page = $('.hfeed .hentry .entry-title'); and it will work across any site with that markup, which is much better than defining our own POSH format specific to a single site. Thanks again for all the detailed feedback and putting up with this long thread. Fiann 2009/11/27 Brian Suda brian.s...@gmail.com: On Fri, Nov 27, 2009 at 2:12 PM, Fiann O'Hagan fia...@jshub.org wrote: What typically happens on big enterprise sites is that they have an analytics product which requires certain per-page metadata, such as a page name and category. --- yup, I know them well. One solution would be to define your own POSH format and/or re-use something like hAtom. Then in the JS code for declaring variables for tracking you can reference the microformats, for instance: Instead of: var page = news-index; var campaign = news you could replace the declared variables with references to the visible text such as: var page = $('.entity-title'); var campaign = $('a[rel=tag]'); In the JS you are referencing visible data. As editors change fields in the CMS, the tracking codes, campaigns, sections, and other tracking is done automatically. What you need is the mapping between the visible parts of the page and your specific tracking variables. It also depends on how much you want to connect the two and/or allow editors to be changing these values. -brian -- brian suda http://suda.co.uk ___ microformats-new mailing list microformats-new@microformats.org http://microformats.org/mailman/listinfo/microformats-new ___ microformats-new mailing list microformats-new@microformats.org http://microformats.org/mailman/listinfo/microformats-new
[uf-new] Microformats for hidden data
Hi everyone, A little while ago my colleague Liam posted on this list about the jsHub project and our ideas for a microformat to replace the proprietary JavaScript currently used for web analytics metadata. He got some good feedback, and I can see there's work we need to do. Here's the use case we want to address: there is a lot of information currently stored in pages which is encoded in vendor-specific JavaScript variables. There are many reasons why the microformat approach (in principle) would be better than the current situation. Publishers of big sites find that they are now using multiple tags, and therefore it makes sense to have a single version of the data about each page rather than re-declaring it in multiple formats. I also believe that some of this information (page name and category, for example) would be of great interest to search engine spiders if it was accessible. I would like to take a step back from comments on our specific proposal and ask a much more general question. Are there any materials currently available about information which is not in the visible HTML of the page? As far as I can see, all the microformats currently in use start with information which is visible in the page, and then add markup to indicate what it represents. For example, with hProduct, you start with the existing product name, price etc in the page, and add the appropriate classes to indicate what these fields represent. But there is a wealth of information hidden within the page in meta tags and in JS blocks. For example on the microformats.org wiki at http://microformats.org/wiki/hcard-faq var wgPageName = hcard-faq; var wgTitle = hcard-faq; var wgAction = view; It's quite possible that for web analytics purposes, you might want to use the page name hcard-faq which is different from both the HTML title element hCard FAQ middot; Microformats Wiki and the URL path /wiki/hcard-faq. Is there any guidance available about these cases, where the information we want to capture is not part of the visible page? Please note that it is human readable, but the person consuming the data is different from the end user browsing the site, for example, it is someone looking at reports on the most popular pages on the site. This means that some of the microformats principles, such as visible data not invisible metadata, can't directly apply. Thanks for any feedback, Fiann ___ microformats-new mailing list microformats-new@microformats.org http://microformats.org/mailman/listinfo/microformats-new
Re: [uf-new] Microformats for hidden data
Thanks for the quick response Dan. You already highlight the existence of meta, and I guess I'd just draw a stronger contrast between that and proprietary / random Javascript variables. There's a lot to be said for not having to run a Javascript interpreter to figure out the basic data structures encoded in a Web page. So maybe meta is worth some more investigation, rather than just listing it as part of the problem...? I agree we shouldn't dismiss the meta tag. The problem as I see it is that the meta tag only allows a simple association of name=value pairs. It doesn't allow any kind of structured data. So you can have a meta tag that gives the author, as recommended in the HTML spec http://www.w3.org/TR/html401/struct/global.html#h-7.4.4 For example, to specify the author of a document, one may use the META element as follows: META name=Author content=Dave Raggett But if you want to use hCard to give contact details for the author, you can't, because it's an opaque string. There's additional complexity with the content of the tag being in an XML attribute rather than a text node too, which complicates the escaping required for the string and means that you cannot include any HTML in the text. As I understand it, these limitations are what led the W3C to create RDF, which is cross-linked from the meta element in the HTML spec. And the complexity of RDF, is of course what led to the rise of microformats. One final serious limitation of the meta element is that it is only valid in the head of a document, and not in the body. With more complex pages, for example tabbed layouts, and content served in via AJAX, there's a good case to associate page metadata with a fragment of the page rather than the entire HTML document. That's not possible unless you can define a wrapper element around the content you are concerned with. Does that make sense, or should I be looking at it again? Fiann ___ microformats-new mailing list microformats-new@microformats.org http://microformats.org/mailman/listinfo/microformats-new
Re: [uf-new] Microformats for hidden data
That's interesting Scott. I am not sure I have understood your point completely, but I'd like to explore it. This isn't very accurate. RDF was not created primarily as a response to HTML's limitations, nor microformats as a response to RDF's complexity. I agree. RDF is not about a limitation of HTML, but it is an attempt to allow data which is too complex to convey in the meta tag alone. I agree, this seems much more in line with RDFa than microformats. To do this in microformats, we'd need to throw out the visible data requirement, and re-interpret all of the other guidelines to no longer presume visible data. And after a lot of work, the result would end up looking a lot like RDFa. Why would it end up looking like RDFa? This is the part I don't understand. RDFa looks like it does because it involves XML namespaces, namespaced values for XML attributes, and URIs. The markup indicates relations between items, where the nature of the relation is defined by resolving a URI. In contrast, microformats simply use some well known class names. If we have an element with the class of hproduct, it describes a product. Inside that, an element with the class of fn is the product name. There is no URI to dereference to understand what is meant by fn. So when you say that it would end up looking like RDFa, do you mean in terms of syntax? Or do you mean in terms of the data being applied as attributes to elements that are otherwise visible, like the about attribute being added to a div? If it's the second one, then I was imaging something much simpler, which looks like any other microformat, but with some or all of the content in a CSS display:none region of the page. That to me still looks like a microformat, not like RDFa. Fiann ___ microformats-new mailing list microformats-new@microformats.org http://microformats.org/mailman/listinfo/microformats-new