2018-12-10T21:54:32+10:30, Will Robertson:
>> My question: is this a good solution/compromise for the issues I described?
>> It seems reasonable to me, but I fear I’m probably biased :-)
>
> I definitely agree that unteasing these definitions is a good idea. In particular I like the idea of generalising the formulae in calculating relative font sizes.

While determining what kind of document elements my class will need, I’ll
try to figure out the best places to put the different font size-related
parameters. I think I may need to give amsmath a good read for guidance.

> Frank did mention to me in Brazil that the relsize package has some quirks in that the font size transformations aren’t always reversible — stepping up then down again could lead to a non-zero difference in font sizes. I agree with him it would be important to avoid this kind of imprecision...

It looks like the relsize package uses dimens to calculate the size changes—
I think that using the l3fp engine would avoid such imprecision, assuming
that the scales involved are not extreme. I should probably test this theory.

> Sorry I don’t have too much more to say than that it sounds like you’re on the right track!

No worries, any input, even small, is helpful!


2018-12-10T16:12:03+01:00, Julien Rivaud:
>>> My question: is this a good solution/compromise for the issues I described?
>>> It seems reasonable to me, but I fear I’m probably biased :-)
>> I definitely agree that unteasing these definitions is a good idea. In particular I like the idea of generalising the formulae in calculating relative font sizes.
>
> I wrote something for that, which I think is orthogonal to relsize. The
> code is at
> https://gitlab.com/frnchfrgg-latex/fontsizes/blob/master/fontsizes.dtx

I agree that it is orthogonal to relsize.

My implementation doesn’t carefully control expansion like yours does.
Instead, I relied on the fact that xtemplate doesn’t expand key values until
\AssignTemplateKeys is used. I don’t know if that makes my code more fragile…

> [ It is not yet on CTAN because the only thing using it is my own class,
> which is not ready for CTAN (and might well never be, since it is more
> or less a mix of reimplementations of article/book.cls, inclusions /
> usage of existing packages with settings that I like, and layout
> settings that I consider better than the default ones). The class is
> "rivbook" (
> https://gitlab.com/frnchfrgg-latex/rivbook/blob/master/rivbook.dtx ) and
> looks like
> https://gitlab.com/frnchfrgg-teach/courses/wikis/uploads/b322aa03b98b30dca2e1ec742c48b709/04-statistiques.pdf
> ]

I like the results of this class, and I think the QR codes that link to a
video are a very useful feature.


Warmly,
Kelly