johanneswilm
commented
10 months ago The existing output format with lots of left/right/top/bottom placement instructions will likely make reflowing difficult. This seems to be meant for recreating an existing design exactly. But is not a good approach if the output is to be reused in other ways
Of course, you may just be able to throw those out if the content is in the right order (document order).
Inline styles (as used here) are not a good idea if one wants to use a coherent style on lots of content. Again - these could be stripped out or maybe even be replaced by classes based on the current content (such as all elements with red background styling get a class called "red-background", etc.).
In my own opinion, inline styles work well if one has some unique content that is only used once. For example to draw an image or similar where different divs are placed in specific places to make out parts of the image.
Using H1-6, paragraphs, etc. instead of spans and divs can have advantages in that it makes the semantic meaning of content clearer when a machine is reading the output. Screen readers, etc. will probably also pick up on that easier. Whether that is a concern depends a bit on whether end users will use the output or whether they will only see a version that has already been printed (after going through Vivliostyle).
Alternatively, you may also be able to achieve the same or a similar result by assigning aria-roles to the dovs/spans.
See
[1] https://www.boia.org/blog/accessibility-tips-using-the-div-and-span-elements
[2] https://developer.mozilla.org/en-US/docs/Web/Accessibility/ARIA/ARIA_Techniques
mrchristian
rqpe
Question: I need to ask for a bit of a sanitys check on how we are combining HTML and CSS for a publishing pipeline we are putting together to demo semantic publicaton of UNFCCC publications. We're at a crossroads and we need some advice.
Tagging to see if you can comment please @johanneswilm and @MurakamiShinyu
We have an HTML file that has been generated from Text and Data Miniing software pyamihtml. The HTML currently only has Divs and Spans generated - see example: 1_4_CMA_3_decis.html. The HTML file was generated as a conversion from this PDF file: 1_4_CMA_3.pdf.
The question is if its a good idea or not to be using classes for the Divs and Spans to generate different HTML tags that we'll also be using as CSS type selectors - such as: sections, H1-6, anchor links, paragraphs, lists, etc?
Using the pyamihtml software we can reprocess the HTML file to add classes to the Divs and Spans, or alternatively we can add HTML tags needed.
Is this just going to be bad practice?
Will we hit problems with Vivliostyle rendering?
Is this a problem for using W3C Publication Manifest, generating various ToC? Or having other software that can use W3C Publication Manifest such as https://www.npmjs.com/package/epubjs-cli
Here you can see a work in progress of the sample publication - CMA 3: FCCC/PA/CMA/2021/10/Add.1 (TEST)
Our end objective is be able to automate the whole pipeline for specific Corpus to transform PDFs into a package of HTML, W3C Publication Manifest, and other resources to create publications, and in addition we will be processing the HTML to add semantic markup and classes.