JSON-LD script elements (data blocks) MUST only contain valid JSON

[BigBlueHat]

The current editor's draft suggests wrapping JSON-LD in a data block with <!-- and --> to avoid some HTML parsing situations. Ultimately, that means that we're suggesting that the contents of an embedded JSON-LD data block is legally invalid JSON-LD...which seems bad...

Consequently, I'd like to propose that we only specify the entity encoding of certain strings within the JSON-LD contents to avoid the potential HTML parsing mishaps described in HTML's Restrictions on contents of script elements section.

Doing that avoids encouraging invalid JSON-LD and removes the additional unique parsing steps for cleaning out the wrapping <!-- and -->.

PR forthcoming.

[gkellogg]

I can support this, but we need to have some rule for dealing with comment inside of strings that could cause DOM problems, specifically <!--, -->, <script and </script. Using either or both of the current escape sequences and character references will keep the content legal JSON, but it will not be interpreted properly unless unescaping or resolving character references is performed.

Surrounding script contents with XML comments is a common practice, but may be obsolete now with HTML5, along with CDATA.

[iherman]

This issue was discussed in a meeting.

• No actions or resolutions

View the transcript

JSON-LD script elements (data blocks) MUST only contain valid JSON
Benjamin Young: #96
Benjamin Young: issue related to #97
Gregg Kellogg: what I pushed up there addresses the different aspects of the HTML script content.
… HTML provides 2 mechanisms: specific backslash escapes in HTML comments, or HTML entities
… Benjamin proposed to remove the comments, to keep it valid JSON, consistently with the declared content-type.
Benjamin Young: it’s similar to injecting HTML tags in a <textarea>, without breaking the textarea.
Rob Sanderson: to be sure i understand, the issue is if you have: <script>{"label": "

[iherman]

This issue was discussed in a meeting.

• RESOLVED: close #97 as fixed with recent merges; other issues raised in #100

View the transcript

4.1. Remove the need for wrapping comment-open/comment-close text in data blocks
Benjamin Young: #97
Gregg Kellogg: most of these are related to PRs just discussed, but not all
Gregg Kellogg: #96
Benjamin Young: current draft talks now about html entities for escaping
Gregg Kellogg: given that it has to be valid json-ld, how do we deal with escaped entities
… leaves us with problem of how to represent <script> tags and comments in json?
… so we could close 97 and leave 100 for further discussion
Benjamin Young: one option is to mark propose closing and discuss more later
Gregg Kellogg: DanBri involved and noted in discussion (of this or another related issue 100)
Proposed resolution: close #97 as fixed with recent merges; other issues raised in #100 (Benjamin Young)
Gregg Kellogg: +1
David Newbury: +1
Tim Cole: +1
Benjamin Young: +1
Ivan Herman: +1
Resolution #2: close #97 as fixed with recent merges; other issues raised in #100

[BigBlueHat]

Think I missed referencing this issue in the resolution we used last week. I believe the current text in https://w3c.github.io/json-ld-syntax/#restrictions-for-contents-of-json-ld-script-elements settles this nicely, fwiw, so closing as I believe it's both fixed and we had consensus to match. 😄