http://babelstone.blogspot.ca/2006/06/rules-for-long-s.html

Accordingly, there may be some italic fonts from the late 18th century that follow the contemporary writing practice of /ß for double-s.

I’ve put a long-s feature in several typefaces, but in the “hist” feature (for when and if that gets supported by layout apps) and as a Stylistic Set.

My present thinking is to employ a double long s (with appropriate ligature in the ligature feature) in the code, rather than long s followed by short s (except at the end of words, of course), as I think if people are going to use this feature, they would like to see as many long-s’s as possible!

Here’s my code:

feature ss01 {
sub s by longs;
sub longs' [f fi] by s;
sub [f f_f] longs' by s;
sub longs' [hyphen quoteright] by s;
ignore sub longs' @smcp_source;
sub longs' by s;
lookup one {
sub s' s by longs;
} one;
}ss01;

And (showing only longs ligatures):

feature liga {
sub longs i by longs_i;
sub s longs i by longs_longs_i;
sub s longs l by longs_longs_l;
sub s longs by longs_longs;
sub longs b by longs_b;
sub longs h by longs_h;
sub longs k by longs_k;
sub longs l by longs_l;
} liga;

Not English and not 1700s, I’m afraid (French, 1568), but interesting for showing both a ‘longs_s’ and a ‘longs_longs’ ligature in the same font. The former occurs mainly (though not exclusively) before ‘i’, and is presumably convenient in that position because the font apparently has no ‘longs_longs_i’ ligature.

I just recently saw an eszet in an English text. It was written by Sir Isaac Newton, so that narrows down the date range somewhat. I'll see if I can track down the example.

Ah, here we are. Page 265, starting from here:

http://rstl.royalsocietypublishing.org/content/1/1-22/263.full.pdf

Dr. John Wallis is actually the author of the passage.

For that Diſcourſe is to be looked upon onely as an Eßay of the general Hypotheſis ; which as to particulars was to be afterwards adjuſted , from a good General Hiſtery of Tides ; which it's manifeſt enough that he had not ; and which is in a great meaſure yet wanting.

@quadibloc: wonderful, thanks!

> It was written by Sir Isaac Newton,
> so that narrows down the date range somewhat.

NB:
p. 267, first documented use of a ;) maybe?

Interesting. I went to ſearch the original for more occurrences of double-s words to see if they are all handled the same, and lo, on page 2 of that same document, the same word "Essay" appears sans eszet in the very first plain line of text.
Seems kind of random to me -- could either be a Compositors' Error?

It's possible that there was a convention that Italics, being more swash-y, used more ligatures than the Roman. Maybe, too, there was no Roman eszet for the typeface in use.

On "Munday, Auguſt 6, 1666," English spelling was not particularly standardized, and it would not be until Samuel Johnson and Noah Webster did their thing. Variations on the part of the compositor were not so much error as license, especially subject to how many ligatures they had in the case.

> not so much error as license

x2

That lack of standardization impresses me again and again, particularly when it comes to handwritten stuff. I'm reading a lot of late-1700s-early-1800s in my dayjob, written by scholars, politicians, officers, scriveners, and they would switch back and forth between different spellings of the same word within a single sentence.

Munday, Auguſt 6, 1666

Great. Now I have to read it again to see if there were any fire insurances on offer.

As John Hudson is fond of pointing out, when to use the long s varies between languages and time periods both; any rules-based approach to automatic substitution will only be good for one particular "standard" for long s implementation and generate "wrong" results for others....

True, but there’s nothing wrong with implementing an /ſ feature for a particular language (English) at a particular time (17th and 18th century), which is what I’ve done.

Ideally, it would be most appropriate for Caslon or Baskerville typefaces, but that kind of precise historical allusion is rare—most typographers prefer to use old style types for 19th century subject matter for instance, rather than didones.

I’ve made other discretionary language-specific features, such as ordinals (English), lowered umlauts (German), and Turkish non-ligatures, so why not an /ſ Stylistic Set?

**

Further to Kevin’s insight on typographic licence, I recently posted this image in another thread, and noticed that /ſ is not employed where an /s ligature is available (C’est, cascatelle), or to start a word (sa)—presumably because by the time the reader gets to the second syllable of baiſée or the third syllable of embraſée it’s quite clear what the word will be, and there is no likelihood of mistaking ſ for f, which might occur with ſa. Light seasoning, as it were.

Verlaine: «À la Princesse Roukine», Parallèlement. Paris, Ambroise Vollard, 1900