- From: Roberto Castaldo <r.castaldo@iol.it>
- Date: Thu, 27 Jan 2005 09:10:14 +0100
- To: "'EOWG'" <w3c-wai-eo@w3.org>
Hi group, Sailesh: -Automated evaluation: Certain accessibility barriers can be detected in an automated fashion across a site efficiently using a tool. The results are very reliable because the criteria for evaluation are specific and objective. Form control has no label associated explicitly checkpoint (12.4) or deprecated tags are used like <u> (11.2 or images or area element have null or blank alt (1.1). Absence of title for a page or frame is another instance that can be auto-detected. So there are several barriers that can be reliably detected in an automated fashion. Roberto: I agree, but only if we're talking about using deprecated tags and other few really automatic evaluations a tool can provide; but let's make attention not to give a wrong message: even this kind (or any other kind) of automation cannot replace the intervention of a skilled human developer, and this must be clearly said in our document if we decide to speak about automated evaluation. Sailesh: -Automated repair: Some repair techniques can be implemented in an automated fashion accurately and reliably. Adding keyboard equivalent events like onFocus and onBlur for onMouseover and onMouseOut is an example. Yes certain barriers can be detected with user interaction and certain barriers can be fixed using user input. Roberto: Does really an automatic repair tool exist? I don't think so. I can see that some limited kind of automatic problem detection software exists, but then, when a developer has to fix the problem, he cannot count on any algorithm-only based. Sailesh: Consider, once an alt text is given for an image, an automated tool can create an alt-text repository and use that alt-textfor all occurrences of the img on the page or site. Example: there is a img of a pdf icon next to every PDF file on the site and it has a null alt. The user enters "PDF Acrobat File" as alt-text and the tool assigns it across the site. Would one characterize this as almost entirely automatic or as semi automatic? True it cannot judge if the alt-text is appropriarte. Roberto: I think this is a semi-automatic one; if a single page contains several images of different PDF files, the procedure above fails, as it gives the same alt-text to different images; i'm trying to say - as Sailesh himself says - that developers must discuss about any alt-text in any page; no tool can correctly perceive context. So, if we consider it as an automatic task we risk to give a completely wrong message. Sailesh: A tool can have a rudimentary algorithm for detecting data tables and might again ask the user for confirmation. Once it is confirmed as a data table, the tool can ppoint out that it has no markup for associating header and data cells. Roberto: Once again the tool can only detect a data table, but the developer has to understand the problem and give his own solution. The software tool can guide developers (and i'm not sure it's always a good thing), but cannot complete the repair task on its own. So, once again, we're talking about semi-automated repair tools. Sailesh: So I think it is unfair to not even recognize the presence of automated tools for eval and repair. The only reasons certain barriers cannot be identified in an automated fashion is because there is no objective criteria for detecting these. Roberto: We could consider to add a sentence or a small section dedicated to automatic evaluation (not repair) tools, and this can be useful to make the document more complete: a not skilled developer reading our document may ask himself: "what about automated evaluation and repair? Why this document doesn't talk about them?". If so, this section should stronlgy warn the user that automated tools can only give some suggestions, and cannot complete any task in repairing a not accessibile web page. My best regards, Roberto Castaldo ------------------------------ Webaccessibile.Org coordinator IWA/HWG Member r.castaldo@iol.it rcastaldo@webaccessibile.org
Received on Thursday, 27 January 2005 08:10:49 UTC