Adding insular script like it's 1626

To the point: that's a neat use of <ruby> in a non-CJK context, amonf other things! And a neat use of discretionary ligatures, of course.

If I feel a need to, I usually use the compose keys to put actual dots ("tá fáilte roṁat, a ċara"), but that's very unfriendly to screen readers and other tools that need to interpret the words and expect the modern spelling with h's, as the author points out..

Related but not really on the technical subject: there are some common misconceptions about the evolution of that script.

It uses a dot over a consonant to indicate it should soften according to grammatical rules.

The author implies the reader understands what it means but I'll elaborate for the uninitiated for my later points to make sense. The story was that at some point in the 4th-5h century, Insular Celtic languages experienced a consistent shift of intervocalic consonants (consonants placed between two vowels) that, simply put, eliminated voiceless plosives. The classical Latin 'sacerdos' /sakerdos/ (priest) changes to something like /sagerdos/. The old Indo-European "equos" (horse) became something like /exos/.

That's by itself is very common, like in French "cheval" where the /k/ of the Latin "caballus" became a "sh" and the "b" became "v".

To know that "cheval" is related to "caballus" one needs to know what the process was, but in Irish, a lot of former consonants are still visible. E.g., the 'bh' in "leabhar" sounds like /w/ but clearly hints that the original form was the Latin "liber".

The fun part is that it happened not only to consonants inside words, but also to consonants that ended up between vowels at word boundaries. For example, "sinda bena" (the likely reconstruction of "the woman") came to be pronounced "seenda veh-na", with the voiced bilabial plosive lenited to a voiced fricative. That became "in ben" in Old Irish — the lenition pronounced but no marked in writing yet, and "an bhean" in Modern Irish — "bh" representing a lenited /b/.

Then a lot of such things became grammaticalized and there are particles that look the same but differ in mutations they cause. For example, "a cat" means "her cat" but "a chat" /a xat/ is "his cat" — the feminine singular possessive pronoun/particle causes no mutation to consonants, the masculine one lenites them.

The outcome is that one cannot "just" create a perfectly phonetic Irish alphabet because lenited consonants come in two varieties: embedded inside words from the time when it was a phonetic process and those that occur due to grammatically significant mutations. Any writing system will have to compromise and sacrifice either the simplicity of representing sounds or the connection with the original, unmutated word.

Another factor is that historical lenited consonants inside words may or may not be silent in different dialects, like the 'gh' at the end of many verb forms is silent in the North and the West but pronounced in the South ('nigh' (wash!) is /ni:g/ in Cork and Kerry but /ni:/ elsewhere.

The “old script” was created when medieval monks evolved the latin script to encode the sound of Irish. The solution was to switch to using a pure latin character set, with the dot being replaced with a “h” after the letter.

The original writing system of Old Irish was a pure Latin script with h's used a modifiers. It was quite a mess: they used 'ph' for the lenited /p/ (sounding /f/), 'ch' for the lenited /k/ (sounding /x/), and 'th' for the lenited /t/ (sounding /θ/) — presumably because they were familiar with that convention from Latin transliteration of Greek words (like "[Iesus] Christos").

Most instances of lenition were unmarked and are understood to have been there on the account of the later forms and the existence of ph/ch/th — phonetic processes like that don't occur piecemeal. They also used double letters to mark unlenited consonants, like in "macc" (son) to show that it was not /mag/.

Then people started using 'ḟ' for the silent lenited 'f', apparently from a scribal convention to use dots to instruct the reader to ignore erroneous letters ("punctum delens"). Then that convention spread to other letters in Middle Irish and the classic "cló Gaelach" was born.

This is why you get Irish names like “Sadhbh” that improbably rhyme with “five”. It’s much less phonetic because you have to look at the preceeding letter to understand what the “h” modifier is going to do, and you have to process letters in groups to understand what sound they make.

The reader will always have to process something — either the H or the dot, and know that lenited D's before other consonants are silent or pronounced similar to /y/ as a rule in Modern Irish.

My favorite example is the word "oíche" (night). It used to be written "oidhche" before the 1940s spelling reform that removed a lot of silent letters. Why did it have the seemingly absurd "dhch" cluster? The Old Irish word for night was "adaig" (the 'd' was like 'th' in 'the' and the 'g' was a velar fricative). Its dative singular form was "adchi", pronounced with a virtual /ə/ vowel between the the two consonants, as it still happens in Modern Irish. At some point in the breakdown of the original case system, dative singular forms started to be used in place of nominative singular forms. At the same time (or maybe a different time), dental fricatives disappeared from Middle Irish and eventually the oridinal 'd' disappeared from the sound of that word completely but remained in its written form for centuries.

Nice hack! You know, now that I think of it, typewriters and to a lesser extent (cheap) movable-type typesetting were a massive evolutionary bottleneck on Latin scripts. (Expensive movable-type shops could probably do whatever you asked them to.) Half of the spacing and line-breaking stuff discussed here is due to that as well. Now that we have computers with high-res bitmap displays, fancy font rendering, and the ability to automatically download and use whatever font a document's author suggests, a lot of the constraints are relaxed and you can get a lot more creative with your hacks.