The discussion revolves around aligning schema.org markup between the Discourse-solved plugin and the core forum platform (DFP) for QAPage. rrlevering highlights inconsistencies in author information, content truncation, answer count, and date fields between the two markup types, which hinder deduplication. sam acknowledges the issue and commits to reviewing it. techAPJ confirms that key issues—particularly author information and date discrepancies—have been resolved via a commit. sam asks if the issue can be closed, and rrlevering agrees, noting the alignment allows for deduplication on their side. He suggests a long-term solution: either native integration of types in Discourse or using node IDs to append types to core markup, acknowledging current practical limitations.
AFAICT from reading other posts, the schema.org markup for QAPage is generated from the discourse-solved plugin. I’ve read Wrong answerCount value from discourse-solved - #6 by sam for instance which has some discussion on problems conforming a more generic forum to the q-a model.
To consumers (including me/Google) this is somewhat duplicated markup with the DFP that is produced by the main forum. In fact, both of them are currently marked as the main entity of the page. Annoyingly they also don’t fully overlap in their contents which make some naive dedupings fail. For instance:
the author of the Question/Answer markup uses the human readable name and no URL, but the author of the DFP uses the handle + URL. This makes the authors look very different.
the content of the Question/Answer markup is truncated, but the content of the DFP because it’s microdata is fully there
As discussed in the aforementioned post, answer count is sort of fabricated
dateCreated is used in Question, but datePublished is used in DFP
The title and date seem to be consistent, but sometimes our deduping logic needs more when it’s implemented globally.
As a minimum improvement without major changes, do you think maybe you could align the author information so that both of them have handle, name (you can use alternateName - Schema.org Property to differentiate), and the profile URL in both places consistently?
Thank you for the fixes. I think the markup is aligned enough (mostly from the author information) that we can do deduplication on our side now using some combination of author and timestamp. To let you know what I plan, I’m going to essentially merge the extra Question type into the DiscussionForumPosting and the Answer type into the Comment on our side.
It would be advisable to eventually figure out a way to do this on your side. Either integrate it directly into Discourse for native typing in the markup to pick between DFP or Question OR use node ids to append the extra type on the markup generated by the core forum platform (so that the OP will be co-typed Question +DiscussionForumPosting). It’s probably eventually going to be disallowed or at least marked invalid in reporting by our systems, but I understand some practical reasons why this is necessary currently as we’re figuring this space out.