any miktex users around? tex.stackexchange.com/q/660239/1090

@DavidCarlisle I just confirmed the issue on the miktex github.

@UlrikeFischer thanks

@DavidCarlisle oh no

@JosephWright thanks I started to make a PR last night but was distracted as I have some unrelated changes in the upload script in a local branch, I need to remember why I didn't push those at the time (last december) I couldn't see open isssues here possibly it was something i'd spotted, or does this look familiar?

## [Unreleased]

### Fixed
- use `form-string` rather than `form` for all curl fields to avoid
  misinterpreting leading `@` or `<` eg a description starting `<p>`

- Check the boolean value returned by executing shell commands in
  l3build-upload and throw an error if this is false. This fixes
  the issue that previously "validation successful" was reported
  if curl failed.

@DavidCarlisle Vaguely: I think it's just a question of going for it

@DavidCarlisle Planning a TUGboat in response to Karl? (I'm going to do two-part tables I think - might be useful to someone like Petr)

@JosephWright yes I'll try to test it and probably make a pr later, testing upload without actully uploading to ctan is always a pain, normally can use the debug option but for that last one and in this local branch I have a modified version that uses real ctan for validation but the postbin test for upload

@JosephWright I wasn't planning anything well I could have done something about \symbf if I'd written the code:-)

@DavidCarlisle :)

@DavidCarlisle More-or-less my plan: write article and finish code so it is right :)

@JosephWright what do you mean by "two-part tables"?

@UlrikeFischer compression technique for unicode tables

@DavidCarlisle That's the one

@DavidCarlisle ah ;-)

@UlrikeFischer Quite technical article, about how best to handle Unicode data at a low-level in TeX

@JosephWright it would be nice to see such an article.

@UlrikeFischer I'll work on it :)

@JosephWright Sounds interesting, indeed!

@DavidCarlisle I dont understand your comment. That does not look like what we see in the question. Was the question perhaps updated?

Ah, you edited it. I did not look at the history of the question

@mickep The question had no example at all, guessed one but I guessed good code, I should have guessed bad:-)

@DavidCarlisle Well, maybe. As it stands now, it is very unclear, since the image probably shows the expected result. :)

(I am not aware if it is common in this area of math to have such a tight spacing around the \leq)

@DavidCarlisle OK, with your updated question and answer, I removed my comment.

@mickep yes idea was to push OP to edit it but as I guessed an answer anyway I have rolled back my change to the question

(Convenient trick to edit the question so that it becomes strange, write an answer that fits the original one, and then to edit the question back. You will get ahead of the italian if you proceed so... +1)

@mickep seems like a plan

@mickep @yo' will be shocked to learn an overleaf user has ignored an error message

@DavidCarlisle Well, nowadays they have no excuse to, do they? :)

@mickep well except the users who can find that option are a subset of the users who already knew how to find the log:-)

@DavidCarlisle Ooh no, that is probably true.

Wait what twitter.com/AFC_GLEN/status/1575942075287158785

Hahaha! Thanks for that laughter.

@mickep -- I've left a comment on exactly that point. Since my knowledge of the meaning or intention is lacking, I've expressed my opinion with an analogy. I think that a bit more space might be okay, but not the full "relation" space.

@barbarabeeton So, maybe a tinymuskip or a pettymuskip would be it. :)

@mickep -- Yes. I'm not sure how to do that automatically without digging.

@barbarabeeton I will have a look in some probability books to see how this is usually done.

@mickep -- Good idea. I looked at the AMS style guide and Math Into Type and found no relevant examples. But it occurs to me that maybe binary operator spacing might be acceptable. But not relation spacing.

@barbarabeeton this is how it can look. The first one is from Annals of Probability, the second from some random book from Springerlink. (Others looked more or less the same.)

2020 to go

@Rmano Done by the end of the year. :)

@mickep -- Okay, but in this one, the shape of the "X" makes this one look particularly odd.

@mickep -- Maybe "unbalanced" is a better description than "odd".

@barbarabeeton Maybe. I think that one is very common. I think we also get too picky by staring closely at the formulas.

@mickep -- While I agree in general, I didn't stare at that one. The extra-wide space just popped right out at me. (However, I'm rather sensitive to such differences. I once flabbergasted one of my nephews, who was looking at a license plate frame that he wanted to buy, by asking him to please first check to see if it would fit on his car. It didn't.)

@barbarabeeton Yes, I agree regarding the spacing. Probably the italic correction for the X.

@barbarabeeton Do you happen to know the story of why the \times has such a large side bearing in cm and lm? Is it only so that the different binary operators should have the same space?

(While looking at this I also realized, probably not new to the oldies, that the width of the glyphs in cm and lm are not the same. In cm one has 777.7XX where I forgot the XX, and in lm (OT) they are rounded to 778.)

@mickep -- Certainly that. A desirable "extension" to TeX, requiring knowledge of the shapes of characters in fonts (which, as far as I know, doesn't exist yet). Something of this sort was discussed in the font group in the ANSI and ISO working groups I participated in in the 1980s. Still in the future, I expect.

@barbarabeeton The closest I know is staircase kerning, but that is of course not knowing the shape.

@mickep -- Recognize the computer memory space Knuth had to work with. He cut down the parameters defined to the minimum that would produce a result he felt was close to the models he chose to follow. So the spacing around the various classes would be the same. Using upright "H" as the "variable" would give a better visual comparison. That's what's used by font designers when they're adjust side bearings. ("Hamburgerfonts")

@mickep -- I have no idea why the metrics for cm and lm are different.

@DavidCarlisle 400 left ^^^

@CarLaTeX 2 days:-)

@DavidCarlisle 2 days of yours, 2 months of mine

@CarLaTeX just make sure no one downvotes you:-)

@DavidCarlisle In case, I can ask a question, accept an answer and then downvote something else :)

@CarLaTeX ooh ask a question: interesting technique

@DavidCarlisle a technique you and egreg can't take advantage of :D

@barbarabeeton I am aware that Knuth had a limited memory space (and just a few different widths). I think that the 778 is rounded 777.7XX. Probably because the OT format only allows integers.

@mickep -- Have a look at this: tug.org/TUGboat/tb03-2/tb06software.pdf I don't remember exactly, but I'm pretty sure that Knuth had to mash everything into just a few meg, maybe only one meg. I know it's stated somewhere, but at the moment I'm not sure where to look. The OT integer restriction sounds logical.

@barbarabeeton Thanks, I'll look tomorrow.