QT4 CG Meeting 154 Minutes 2026-03-03

• [X] David J Birnbaum (DB)
• [X] Reece Dunn (RD)
• [X] Christian Grün (CG)
• [X] Joel Kalvesmaki (JK)
• [X] Michael Kay (MK)
• [ ] Juri Leino (JLO)
• [X] John Lumley (JWL)
• [X] Alan Painter (AP
• [X] Wendell Piez (WP)
• [X] Norm Tovey-Walsh (NW) Scribe. Chair.

Proposal: Accept the agenda.

Accepted.

Proposal: Accept the minutes of the previous meeting.

Accepted.

The next meeting is planned for 10 March 2026.

Note: Daylight Saving Time begins in North America on Sunday, 8 March. It does not begin in the UK and Europe until Sunday, 29 March. The Invisible XML CG meets on UK and European civil time. The meetings of 10, 17, and 24 March will be one hour later in North America.

• [ ] QT4CG-143-02: MK to try to recover the ability to extract formal equivalences into tests
• [ ] QT4CG-143-03: JK to look for C14N test suites.
• [ ] QT4CG-144-01: MK to consider if any now lost value comparisons should be added as examples.
• [ ] QT4CG-149-01: CG to draft a new PR with the common rule note.
• [ ] QT4CG-150-01: NW to ask Jirka for a room at XML Prague for Tuesday/Wednesday
• [ ] QT4CG-150-04: NW to see about a status update on PR #2345; possibly schedule discussion
• [ ] QT4CG-151-01: MK to fix the typo in the build-dateTime function.
• [ ] QT4CG-151-03: NW to investigate why records are formatted incorrectly in XSLT
• [ ] QT4CG-152-01: MK to add a note about the new and old syntax for names in PR #2446
• [ ] QT4CG-153-01: MK to investigate the indentation problem.
• [ ] QT4CG-153-02: MK to turn my suggestions on #1949 into a PR.

This section summarizes all of the issues and pull requests that need to be resolved before we can finish. See Technical Agenda below for the focus of this meeting.

See PR #2467

I think the high-order question here is, do we want to add fallback to the text functions, or leave them alone and add it to bin:decode-string?

NW attempts to outline the situation.

• MK: It’s generally true that if you use off-the-shelf decoding libraries, it’s hard to catch decoding errors.
• CG: I agree. It could be hard to do this properly.
• NW: There are three options…no, only on text, or on both text and ~bin:decode-string~f
• CG: We have permitted characters now.
• RD: Should we do what decode libraries do and insert a fallback character?
• MK: Is it meaningful to distinguish failure to decode from having a character that isn’t permitted?
• NW: I think we’re talking about unpermitted characters.
• JK: One other model to consider is instead of using a fallback function using a filter.

Straw poll: fallback function or fallback character?

In favor of a function: 1, in favor of a character: 2.

• RD: It appears that ICU allows a callback. https://unicode-org.github.io/icu/userguide/conversion/converters.html#error-callbacks
  • … We should do some more investigation.
• NW: Let’s stop there and I’ll see what I can do about some research. This isn’t worth spending the whole call on.

See PR #2470

• MK: This proposal does two things which might have been a mistake based on comments.
  • … In effect there’s a new function “matching substrings” that gives you more detail about the things that you matched.
• ML: The other thing it does is to make all the regex functions available as methods on a regex object; there’s a function fn:regex that returns a record with functions as its members. That allows you to call matches, replace, etc. as “methods”.
  • … That makes it easier for the user to indicate that the compiled expression should be cached for reuse.
• MK: I’m interested in discussion; we should probably separate those.
• NW: I’m a little worried about the proliferation of using records as pseudo-objects.
• CG: I think adding the regex object is a bit redundant and may confuse users.
  • … Regex isn’t too complex so you can use things like an LRU cache of the compiled objects.
  • … It seems to me that there are harder things in the language to cache.
• JK: The motivating user story came from Liam Quin. Any word from him?
• MK: I think he indicated that matching-substrings exceeded his requirements.
• CG: I really like that function.
• MK: What I’ll do is progress matching-substrings as a free standing function and I’ll consider making fn:regex a separate proposal.
• WP: Does this fall into a class of such problems that are solvable by record structures?
• RD: The issue is caching the parsed regex object.

See PR #2481

• MK: I think this is fairly straightforward. It reduces the number of keywords we reserve.
  • … The list basically includes all the spelled-out binary operators.
  • … It includes the leading keywords of clauses in flower expressions.
  • … And a couple of odd ones like case and else that are kind of pseudo-operators.
  • … Any keyword that can appear between two expressions becomes ambiguous.
• RD: That’s because in those case the right hand side will be an empty map.
• MK: Yes, the { introduces a map or empty constructor.

Proposal: accep this PR.

Accepted.

See PR #2483

• MK: This ended up being a very simple editorial proposal. I was thinking of doing more but came to the conclusion that the text is largely there.
• MK: Where we talk about namespace declaration attributes, it presents the rules in a more structured way. There aren’t any semantic changes.

Proposal: accept this PR.

Accepted.

See PR #2490

CG outlines the issue.

• CG: In discussion with users, it seems like JNodes are more complicated than they are.
  • … We can rename some of the functions to make them clearer.
• RD: I like this. One of the advantages this has is that it reuses the names from maps.
• JWL: Is there a mechanism in the system we have that allows me to move back up from an XNode up to the root and then further up into a map or array.
• CG: Currently, no. The ancestors of XML nodes end at the XML root.
• JWL: I’m thinking of the case where the value of an array is an XML tree.
  • … I’ll have to think about it.

Proposal: accept this PR.

Accepted.

See PR #2491

MK introduces the PR.

• MK: This recognizes that the type node is equivalent to a choice of one of the seven node kinds.
  • … It generalizes to the other choice item types.
• MK: This picks up a few cases where it’s clearly equivalent and the subtyping rules now recognize that.
• RD: In the data model spec where we have the hierarchy diagrams have we included JNode and GNode in those?
• MK: Yes, that’s already done.
• RD: And those reflect these changes?
• MK: I’ve changed anyAtomicType into a square box to indicates it’s an abstract type.

MK reviews the diagram in the Data Model spec.

ACTION: QT4CG-154-01: MK to review why anyAtomicType didn’t, in fact, become a square box!

Proposal: accept this PR.

Accepted.

See PR #2492

• CG: From the rules, it wasn’t clear to me what was supposed to be done with the $usage parameter.
• RD: If you do specify a collation, it can be used anywhere?
• MK: No, some can’t be used for substring matching for example.
• RD: Isn’t that what $usage was for?
• MK: Yes, but it’s not clear that things are as clear cut as the argument proposed.
• RD: Would it then make sense to take a sequence of function references?
• MK: But it’s used in other contexts too.

Proposal: accept this PR.

Accepted.

See PR #2474

• MK: The problem here is “what exactly do package parameters mean, and when can you set them?”
  • … How do the answers impact what it means to pre-compile a package?
• MK: We now allow xsl:with-param as a child of xsl:use-package
• MK: There’s a new section 3.4.4 Supplying Package Parameters
  • … Use package parameters must be known statically.
  • … One could generalize it to allow global variables, but I decided not to.
  • … It has to be a select attribute, not a sequence constructor
  • … It’s not an error to provide a parameter that isn’t used.
  • … You can set static or non-static parameters
    • … But observe that some values may already be fixed from some earlier part of the compilation; that raises an error.
• MK: There’s an existing mechanism that you can override a parameter that’s still allowed.
• MK: There’s a little extra text on xsl:expose to tidy up some issues with visibility.
• JK: I like being able to control the parameters this way.
  • … Is it worth a note that this isn’t a “tunnel” parameter; they don’t go down to descendant packages.

ACTION: QT4CG-154-02: MK to add a note about “tunneling” package parameters.

• JWL: Can you do anything with the override mechanism that you can’t do with the new functionality, or vice-versa?
• MK: There were very old statements about whether stylesheet parameters were public or private. I’ve clarified that they’re private. That means you can’t override a parameter with a variable.
  • … What you can do, but isn’t very useful, is override a variable with a parameter.
  • … If the use-package declares a global variable, you can override it with a parameter that can be externally set.
• JWL: It might be worth contrasting them in the spec.

Proposal: accept this PR.

Accepted.

See PR #2478

ACTION: QT4CG-154-03: NW to make sure that the streaming spec goes in the diff list

Leave this until that’s fixed.

See PR #2479

• MK: The term “potential visibility” was a hangover from an earlier draft.
  • … It was used in two places. It obviously meant “declared visibility” in the figure caption
  • … And it was used as a loose English adjective.

Proposal: accept this PR.

Accepted.