Character entity larrpl isn't left-arrow-with-plus, to match rarrpl

[ediosyncratic]

Either of these pages:

• https://html.spec.whatwg.org/multipage/named-characters.html#named-character-references
• https://www.w3.org/TR/html5/syntax.html#named-character-references

tells me:

• ⥅ is U+02945; this is ⥅ RIGHTWARDS ARROW WITH PLUS BELOW
• ⤹ is U+02939; this is ⤹ LEFT-SIDE ARC ANTICLOCKWISE ARROW

Compare U+02946: this is ⥆ LEFTWARDS ARROW WITH PLUS BELOW
This seems a far better fit for ⤹ !

[ediosyncratic]

The table of character reference names appears to be generated from a file, entities.inc or entities.json, which I do not find in the source tree, so I am unable to offer a patch - unless someone cares to tell me where to find that file.

[domenic]

This isn't something we can fix; it's just how these entities were historically defined. Try them in your browser! ⥅ ⤹ yields ⥅ ⤹ indeed.

[ediosyncratic]

@domenic: close inspection of the source will show you that I used them in the bug report; so yes, I know browsers currently support the present broken spec; however, the thing that would fix it is precisely if this spec recognised a clear mistake and introduced a backwards-incompatible change. I sincerely doubt more folk use ⤹ deliberately for its specified meaning than use it expecting the meaning I suggest (and I'm fairly sure both groups are small, probably empty); and the browsers aren't hand-coding this stuff, they're surely using the entities.json file or something derived from it to generate their entity mappings, so a fix by WhatWG would soon enough fix all browsers.

[domenic]

I see. Well, I'll tag this as "needs implementer interest", but I'm not sure how many browsers are going to be interested in changing the meaning of existing pages just to make entity names a bit more symmetrical.

[ediosyncratic]

It's not just a matter of being "a bit more symmetrical" - there's a pattern in the naming that leads the reasonable reader to suppose ⤹ means "left arrow (with) plus"; but the spec wantonly has it mean something completely different.

As it happens, the given meaning for larrpl is the mirror image of what's given for cudarrl, U+02938; whose mirror image should clearly be cudarrr (U+02935, ARROW POINTING RIGHTWARDS THEN CURVING DOWNWARDS; apparently there is no matching left-then-down arrow). These are another pair whose names are incoherent; they should clearly be each other's mirror images, but both curve clockwise; cudarrr starts out pointing right and curves down; while cudarrl points down, initially a little to the right but curving to the left. One could justify either by a cunning reading of its name (cudarrl points down and curves left; cudarrr curves down from a rightwards start) but you need incompatible readings to make sense of them. (Contrast with the symmetry between rdca ⤷ and ldca ⤶, for all that their pattern implies a family, the rest of which is missing.) Still, this one really is just a borked symmetry (with incoherent naming); while it probably would make sense to change cudarrr to U+02939, this isn't as clear an error as the larrpl one, for which ⥆ is unequivocally the only sane meaning.

It may be worth checking ISO 8879 (I don't have access; there's always a pay-wall in the way), which is a source document for at least some of this mapping (e.g. the heinous there4 ∴ for which &thus; or &so; would have been terser and so much more apt). If the present situation is inherited from a mistake in it, then you really can give up on fixing this; conversely, if the present situation is a mis-copying from ISO 8879, that would strengthen the case for fixing it.

[annevk]

Changes to the HTML parser need to meet a very high bar (basically only security issues, non-normative changes, and <template> have succeeded since things settled down). Unfortunate as it is, I think this proposal is made about a decade too late.