- From: Jason White <jasonw@ariel.ucs.unimelb.EDU.AU>
- Date: Sun, 5 Oct 1997 12:21:11 +1000 (AEST)
- To: WAI HC Working Group <w3c-wai-hc@w3.org>
It may be necessary to define additional ACSS properties to control the reading of tables, which could be incorporated into CSS 2. These should include the ability to specify the order in which the cells in each row should be read, and any text which is to be announced before or after the contents of each data cell. Thus, to expand T.V. Raman's example, one might envisage a two-column table, with the first column designating the year, and the second giving the corresponding human population growth (as a number of persons). A style property could then be defined which would take a value similar to the following: "Population growth in {col=1} was {col=2}" The cell designation is thus embedded within the text, enclosed in braces as shown. More complex cell coordinates should be expressable, and of course it should also be possible just to specify a reading order without any surrounding text, as in: "{col=2,h} {col=1,h}" where the "h" suffix indicates that the corresponding header cell is to be read prior to each cell. The table would thus be read row by row, with an announcement of each column header prior to the reading of the associated data cell. Note that the foregoing examples make use of a syntax which differs from that offered by CSS 2. I have represented the reading order as a string which may contain text as well as embedded commands that refer to cells in the table. Perhaps it would be better to take advantage of a coordinate system (with variables that stand for the "current row" or the "current column"), than to employ commands of the type exemplified above. If a coordinate system were used, how would one handle cells which span rows and/or columns, and still retain some degree of generality in the style that specifies the reading order, so that it need not be rewritten for each particular table? The style property which I am proposing would take the form: reading: STRING where STRING may contain text and/or embedded commands as previously described. The user agent would interpret any text enclosed in braces as a command that refers to one or more elements of the table, according to a command syntax that would need to be defined.
Received on Saturday, 4 October 1997 22:21:33 UTC