Re: Alternative music font

[Simon Tatham]

(I hope this reply to the list works. I had to post my previous
message through the Gmane interface, but if I have to post this one
the same way, I won't be able to get the In-Reply-To header to work
properly.)

Jan Nieuwenhuizen <address@hidden> wrote:

> Wow.  You created a full font?  That must have taken quite some time!
> I think Feta took Han-Wen and me something between one and two
> man-years of work.

This one has only taken me a couple of months (including some
initial thought about how to get nice-looking curves without an
excessive amount of manual specification). But then, it's very
likely that a lot of yours is better thought out in many ways that I
didn't pay much attention to. (Just for a start, I haven't
implemented your subtle variation between the different point sizes,
except in the braces.)

> I feel a bit disappointed because one of my goals was to create a font
> that would look like the most beautiful music that I have seen.  As
> one of our explicit goals for LilyPond is for the printed music /not/
> to distract the player, we evidently failed to achieve this for you.

I'm afraid so, but then, it doesn't seem surprising to me that one
answer doesn't satisfy everybody's tastes! I don't think you have
any call to feel disappointed at not having managed to please
absolutely everybody.

> Looking at Gonville it's not so difficult to imagine for me how this
> could be, as I cannot remember ever having seen music that looks much
> like it.  For example, the up-flags are much fatter and
> rounder/shorter than the down flags, is that intentional?

I may yet make another attempt at redesigning the multiple flags.
The intention was to have them all essentially similar in shape
(unlike, say, Feta's quadruple down-flag in which the four flags
look very different from each other) and bold enough to make it easy
to see how many of them there were. They're all currently 'the same
thickness' in the sense that every flag covers the same vertical
length of stem where it joins on to it; that's something that I may
re-think later on in favour of a more subjective idea of 'sameness',
because I've already had one mild criticism of it.

> What is the status of the font, is it ready for general use, is it
> finished?

Initial development is complete. I may make changes, but probably
not until I've collected some feedback and got a general idea of
what really does want changing and what's a silly idea I've
accidentally talked myself into by thinking too hard about it...

> Up till now we have been advertising Feta as being "the" lilypond font
> and describing it mostly with general terms as "beautiful" and
> "designed after the best typesetting traditions".  In some places,
> possibly the essay and talks, we elaborated on the fatness, eg see the
> short note of font design at
> 
>     http://lilypond.org/web/about/automated-engraving/typography-features

One comment from a friend about the difference between the two fonts
was that a thing he liked about Gonville was that it looked more
modern. Feta certainly seems to be striving after a 'traditional'
look, and perhaps that's precisely what is not to everyone's taste
(one person's 'traditional' is another's 'old-fashioned' :-).

> Now that you created a second working font for Lily, it would be
> nice if both fonts were [more explicitly] advertised as to what
> they were designed after.  The LilyPond font sources contain
> quite a few citings of sources of inspiration, eg [...]

Sadly I don't have anything like that sort of detailed citation
available. I grew up playing the violin, and in designing Gonville I
was trying to recall the look of the sheet music I was provided with
by my teachers, because that was what I was used to reading;
unfortunately, I don't have most of that sheet music any more, so
all I can give is vague generalities.

Ultimately, my design criterion was that it should satisfy my
personal subjective aesthetic criteria. Feedback so far suggests
that at least a few other people's criteria are not too far off
mine, but I don't think I could really give a scholarly analysis of
where mine came from.

>     Further, common [text-]font considerations were taken into
>     account.  For example, a glyph should look balanced out.  It
>     should not lean backward of forward, inviting the reader to catch
>     it before it falls over :-)

It's interesting that you should mention that: that actually reminds
me of one of my specific issues with Feta, namely that the curved
centre line of its treble clef _does_ make it look to me as if it's
leaning over backwards. Gonville's straight-backed version feels
much more balanced to me.

> Do you intend to have Gonville included in LilyPond?

You'd be welcome to include it if you wanted to, but I hadn't
particularly expected that you would - I was under no illusions that
you'd instantly prefer it to the font you've carefully tuned to the
criteria you consider important! I'm perfectly happy to maintain it
as a third-party accessory, and keep it up to date as necessary. I
don't even ask for a link from the website, if you don't think
Gonville is of sufficiently high quality to merit it.

All I'd suggest is trivial changes to Lilypond to make it easy to
use an alternative font, and at least not actually _deny_ that such
a thing exists. (E.g. the documentation for ly:system-font-load
currently says that only Emmentaler and Aybabtu contain the
necessary LILC, LILF and LILY tables, which is now out of date :-)

Oh, and there was one other thing: Gonville's time signature digits
are deliberately designed to stay clear of the 1st, 3rd and 5th
stave lines, because I find that makes the numbers much more
legible. However, Lilypond insisted on squashing both sets of digits
firmly up against the middle line, defeating my intention. I've
worked around that for the moment by lying about the digits'
vertical extent, but that's a nasty hack; it would be nicer if
Lilypond itself could centre the digits around the 2nd and 4th lines
of the stave in the case where they're smaller than 2*staff_spacing
(which I think would also look nicer in the case where a user
manually selects a small alternative font for the digits). Would a
patch implementing that (which shouldn't change the current
behaviour for anyone using Feta) be likely to be accepted?

> What bothers me a bit is the lightness of the font.  I consider this
> to be an error frequently made by most post-manual/engraver [read:
> computer/programmer] produced music.

Hmm. It probably wouldn't be too difficult for me to redo the whole
thing with increased line thickness; probably not as easy as it
would be for Feta (my metafont-like setup is less highly developed,
which is only to be expected given how much more time you've
spent!), but doable. I could try it and see how it looks.

> I think the glyph lookup and handling code is already parametrized.

Indeed, I was pleasantly surprised at how easy it was to get a font
with different metrics to work at all. Lilypond's strategy of
storing in the font file itself the information about where to
attach stems to the note heads, for example, was extremely useful;
without it I would probably have had to distribute Gonville in a
form that included a source code patch to Lilypond. Instead I was
able to use the unmodified Lilypond binary, and all I had to change
was the fonts subdirectory of $LILYPOND_DATADIR.

The only problem was that I couldn't find any easier way to change
the font throughout than by using an alternative $LILYPOND_DATADIR.
I don't think I managed to get ly:system-font-load to even load a
font at all from outside the data directory, and even if I had, it
wouldn't have been at all clear (if possible at all) how to get
Lilypond to default to taking _all_ its glyphs from that font
instead of Feta. Perhaps I could have reconfigured them all one by
one, but that would need a lot of fiddly Scheme that was different
for every release...

Cheers,
Simon
-- 
Simon Tatham         "I'm going to pull his head off. Ear by ear."
<address@hidden>                          - a games teacher