SEO for LO online help?

Please send LO-related email to the list, not to me personally. Thanks!
Jean

Hi Claire

Search Engine Optimization is close to black magic and a moving target
exercise that Google and others SE lay to prevent frauds and unfair
behaviour.

At the moment, each release of the online Help pages have thousands of
web pages per language and a few basic and proven techniques for SEO,
notably sitemap of all pages and Schema.org that you suggest.

The work was done in 2017 when the new online help was designed. At that
time, I implemented

<body itemscope="true" itemtype="http://schema.org/TechArticle">

<h1 itemprop="articleSection" id="hd_id3153884">The LibreOffice Help</h1>

for all the Help. However I couldn't measure if the recipe worked
because SEs have their own pace for indexing the pages and our volume
isn't so appreciated by them.

Furthermore, our Help system is historically a mix of Reference contents
and User Guide contents, which makes the selection of the right schema a
challenge. Solutions exist but need a bit more work to improve the
internal XML of our Help, such that pages can have the right indication
of its semantics.

So the use of Schema.org looks simple at first glance but it brings
several issues to our current Help implementation.

- The video shows a very simple HTML editing that works with a dozen of
pages. When it comes to 1000's times 50 languages it becomes impractical
to do it manually.

- LibreOffice has its Help pages in Extensible markup Language (XML) to
simplify Help editing, reuse of contents and be translation friendly. As
said above not every page will follow the same schema, and the
transformation from XML to HTML must therefore consider the semantics.
This is not yet implemented.

- Each change in paragraphs attributes such as "itemprop" example above
in the source XML will trigger retranslation of the paragraph contents
wrapped by the <h1></h1> tags, a job our translators often complain.

Sorry for the techy lines above, the idea behind Schema.org is
interesting and is implemented since 2017, but it is still far from
optimal and I'd like to have some assurance that the energy to spend in
this technique worth the effort.

Kind regards
Olivier