> using "Palatino" falls back to "Arial" for the ff ligature

You are using hardcoded ligatures instead of OpenType features? Why?

Yes, sounds like a bug in PrinceXML to me! Although I guess I can see an argument the other way. :/

Thomas: I use OpenType features in my stylesheet. It seems that PrinceXML (the PDF generator) falls back to Arial; perhaps I should contact those guys and ask why it's doing that. I would have expected that, if that ligature doesn't exist in the font file, it wouldn't be emitted...

Thomas: After some digging around in the generated PDF document it turned out that there was indeed a single hardcoded ligature in the text! Removed it, and everything works just fine now, no finger-pointing at PrinceXML this time :-)

Nick is right with asking: "So to which shape should the redesigner/digitizer be faithful?"

Palatino was originally a hot metal typeface. When it was adopted for photo typesetting, the typeface was considerably modified. While hot metal Palatino comes in distinct, specially designed versions for each size, photo Palatino comes only in one design, which is used for all sizes. Additionally, some of Palatino's personality was ironed out; it seems Linotype wanted the new phototype Palatino to be more neutral and less lively.

The digital versions of Palatino available today from Linotype and Adobe are based on the phototype variant. Berthold has prepared a digital version based on the old hot metal typeface, but it is no longer available, as Berthold ceased to exist.

Here you can see a comparison of the digital Palatino from Linotype, which is based on the phototype variant, (top), and the digital Palatino from Berthold, which is based on the hot metal typeface, (bottom):

I much prefer Bethold's digitization, and I think it is a pity that it is no longer available.

In Ulrich Stiehl's Palatino Berthold 1992 BQ pdf document, we can see two versions of the italic, BQ Italic and BQ Werk Italic; do you know the origin of these two italics?

"Palatino was originally a hot metal typeface."

Not in my opinion. The typeface design is based on carved and written forms, originally drawn, and then massively modified in the making of the last generation of metal fonts, by the last sputtering generation of metal font makers.

Oh Berthold's is very pretty! :) These guys seem to have a version to render it but I couldn't find a link to purchase or download the files.

For digitizations of the 1950 Palatino, the typophile link https://typography.guru/forums/topic/75198-forwarding provides interesting information.

is there a tool that helps me compare two types?

You can use FontForge. You open the two fonts and then select "Element > Compare Fonts..."; you get a report on various differences as well as glyphs put in the background for comparison. I tried with Palatino Linotype (from Microsoft Office 2011) and the Palatino from Apple. For letters used in English, the differences are hardly perceptible. As soon as there are diacritics, the differences become more obvious. Here are odieresis and udieresis (Linotpype is filled in green, Apple is only the contours).

And here is a comparison of eogonek

For reference about the expected position of ogonek, cf http://www.twardoch.com/download/polishhowto/ogonek.html

Michael: Thank you so much for the last two posts, that was really interesting! I followed your link and then another to the PDF document set in Berthold Palatino (Palatino BQ). From that PDF and the postscript documents I could extract the Berthold Palatino and... jolly gee does that implementation look warm! :-)

Unfortunately, the font files extracted from the PDF seem somewhat incomplete wrt diacritics and other glyphs.

Unfortunately, the font files extracted from the PDF seem somewhat incomplete wrt diacritics and other glyphs.

Those fonts date back from a period where the number of glyphs per font was rather limited. I have a FontLab demo version and here is a trace of execution in the folder '/library/Application Support/FontLab/Encoding/T1 Roman-Western' of my Macintosh (where one character per line is listed and commented lines start with a %):

% for i in *.enc
> do echo -n $i ; grep -v '^%' $i | wc -l
> done
adobe_default.enc     230
adobe_std.enc     149
iso_latin1.enc     197
mac_roman.enc     256
win_1252.enc     224
%

Even if the maximum number of characters in Type 1 fonts was 256 (including .notdef), as you can see many encodings were far from providing so many characters. I count 202 glyphs in PalatinoBQ-Roman.pfa, you add the "text figures" and smallcaps that are in the Expert font and you get a non negligible number of glyphs for that period in time. Those fonts are from 1992. The first Truetype fonts were released in May 1991 on Mac OS 7.

And, according http://support.apple.com/kb/TA21654, the Truetype version of the new Palatino appeared on Mac OS 7.1 which, according to http://en.wikipedia.org/wiki/System_7, would be August 1992.

Good stuff, thank you Michael :-)

"Palatino was originally a hot metal typeface."

Nope, Hermann Zapf originally designed Palatino for the Stempel typefoundry as a handsetting typeface. An that’s just ‘metal’ ;–) ‘Hot’ metal refers to the line casters by Linotype and Intertype as well as to the Monotype machines. These machines produce set type by producing it directly from matrices in the desired type size. With hand setting, metal type is pre-produced and definitely ‘cold’ when the type setter picks them from the type case. The first Linotype-version of Palatino was called Aldus, a slightly modified version of Palatino, taking the technical restrictions of the Linotype line casting machine in account. Main problem to be tackled: duplex matrices which caused the regular, the italic and the bold weight to be designed at the same widths. https://www.flickr.com/photos/pietschreuders/8389641634/

The Berthold version as mentioned above, is based on the metal / hand setting version by Stempel, not on some ‘hot’ metal / line caster version.