The problem before us is this: Should original double ss be allowed to remain at the end of words, or should it be simplified to s? In the Etymologies, as well as in later sources, Tolkien's spelling is sometimes less than consistent. Consider one word for "joy": In the entry GALÁS, it is spelt glas. Yes this s must originally have been a double ss (otherwise it would have disappeared in this position), and this is confirmed by the Quenya cognate alassë. Tolkien did not list the primitive original that both words are descended from, but it is clearly meant to be *galassê. And indeed, if we turn to the entry BOR, the same word for "joy" is mentioned again – but here it is spelt glass with the original double ss intact! So which spelling should the poor Sindarin lexicographer adopt?
Closely related is the problem of how one should spell such double-ss words when they appear in compounds. In the entry BOR, the noun glass "joy" (so spelt) is mentioned in connection with the name Borlas, of which it forms the second element. The g disappears because of the normal lenition, but notice that here, the double ss is simplified to s. So even if we were to adopt the rule that a final double ss is to be maintained when a monosyllabic word appears by itself, perhaps we should introduce an additional rule to the effect that it is simplified to a single s when it appears as the final element of a compound? Again, Tolkien's spelling is less than consistent. In the entry BES, he mentions bess "woman" as the descendant of primitive bessê; here the double ss is maintained in the spelling of the descendant word. Still in the entry BES, Tolkien also cites the compound herves "wife". This combines bess (lenited vess) with a prefixed element her- that is derived from the stem KHER "rule, govern" - a herves being, etymologically, a "woman who rules [over a household]", paralleling hervenn "husband". Notice that in herves, the double ss of bess has been simplified to s. So far, Tolkien follows the same principle as in the name Borlas, the final element of which represents glass. Yet if we turn to the entry KHER, the word for "wife" is mentioned again, but now it is spelt hervess with no simplification of the final -ss!
Now these discrepancies do not really represent an actual problem - not even in "internal" terms: In a pseudo-medieval world like Middle-earth, with no central language academies to define a "correct" spelling, such inconsistencies would certainly be common. Yet Sindarin lexicographers of our own age should probably impose a consistent spelling instead of mindlessly copying the primary sources. There are two problems here: 1) Should final -ss in monosyllabic words be simplified to -s or not, and 2) even if we do adopt this double-ss spelling when the words occur by themselves, should the final -ss be simplified to -s when the word appears as the second element of a compound? Bess or *bes, hervess or herves? We will deal with these problems separately.
We would normally look to LotR and Tolkien's post-LotR writings to find some genuinely Sindarin examples we could regularize the "Noldorin" material by. However, these sources are sometimes less than helpful in this matter. The Sindarin cognate of Quenya lassë "leaf" is cited in Letters:282 as "las(s)"! Here, Tolkien seems to be telling us that it doesn't matter whether the word is spelt las or lass. However, in the Etymologies, the corresponding "Noldorin" word had unhesitatingly been given as lhass (entry LAS1), and in most post-LotR sources, Tolkien does maintain ss at the end of monosyllabic words:
These examples are in accord with the rule implicitly stated in LotR Appendix E: "Note that consonants written twice, as tt, ll, ss, nn, represent long or 'double' consonants. At the end of words of more than one syllable these were usually shortened." This implies that at the end of words of only one syllable, the long consonants were not shortened - at least not orthographically.
Against this rule and the examples listed above may be cited a few spellings like nos "kindred, family", mentioned in PM:320 as the cognate of Quenya nossë: we might have expected the spelling noss (which actually occurs in a pre-LotR source). Despite this discrepant example, we may conclude that the preferred Sindarin spelling is to maintain double -ss at the end of monosyllables. Actually, this is also the spelling Tolkien normally employed in the Etymologies, when Sindarin was still "Noldorin":
The examples listed above definitely outnumber the words where an original double -ss has been reduced to -s. As we have seen, Tolkien in the case of glas/glass "joy" mentioned both alternatives (GALÁS vs. BOR); in light of all the examples listed above, Sindarin lexicographers should probably adopt the spelling glass. I suspect that Tolkien listed both alternatives in the entry RIS as well, where the published text reads "*risse-: N rhis, rhess a ravine". If rhess is a misreading for *rhiss (the only word that primitive risse- would be capable of yielding), the shorter form rhis would be merely an alternative spelling of the same word. Anyhow, Sindarin lexicographers can safely adopt the form *riss as a word for "ravine" (taking into account the rh- > r- revision as well). If we were to accept rhess as a correct transcription of Tolkien's manuscript, we must presuppose another primitive form as well (probably *rissâ). The Sindarin form and spelling should then be *ress, this being a mere synonym of *riss.
In the entry DYEL, the word delos "abhorrence" is explained as being probably a combination of del and gos (-os), the former representing the root DYEL "feel fear", the latter the root GOS "dread". However, I don't think this should be taken as implying that a word ?gos appeared by itself in "Noldorin" (there is certainly no evidence for ?del as a separate word either!), so we don't have to consider whether this ?gos should be spelt *goss or not. (In the entry GOS, Tolkien asterisked *Goss as the hypothetical "Noldorin" cognate of Quenya Ossë, the name of the Maia: The asterisk indicates that this name - apparently meaning "Dread" - did not appear in this form in "Noldorin".)
In at least one word, Tolkien simplified the spelling of -ss to -s without mentioning the alternative form in -ss anywhere. I think this is a virtually unique example (except for the not entirely comparable cases rhîs, dîs discussed below). The word in question is gas "hole, gap". The entry GAS in Etym lists both the primitive form gassâ and the Quenya cognate assa, and considering all the examples listed above, I don't think Sindarin lexicographers should hesitate to adopt the spelling *gass instead.
NOTE: The form hmas "soil, stain" is very strange. Should it have final -ss? In the entry for the root SMAG, Tolkien first derived the words maw "soil, stain" and mael "stained". Then he had the "insight" that a root in SM- ought to yield "Noldorin" words in hm- (i.e., unvoiced m) instead. Therefore, according to the text as printed in LR:386, he changed maw and mael to hmas and hmael, respectively. Of course, we would expect maw to be emended to *hmaw, not hmas! The latter form would be difficult to account for, given what we think we know about how "Noldorin"/Sindarin developed from the primitive Elvish language. Given the proximity of W and S on a keyboard, I think we can assume that "hmas" is simply a typo for *hmaw, and therefore irrelevant for our discussion here. Incidentally, the Sindarin form should be maw once again, since the revision m > hm was apparently undone later.
In the word rhîs "queen" (RIG), a primitive double ss was probably present at one stage (*rîgisse), but here the simplification to -s is justified by the preceding long vowel: The (rather cumbersome) spelling *rhîss would suggest a super-long syllable which is not actually present. This word can be accepted into Sindarin as it is, except for the revision rh- > r-. The word dîs "bride" can also be accepted in this form. In the entry NDIS in Etym, Tolkien derives dîs from primitive ndîse via "Old Noldorin" ndîs, but since this should regularly have produced "Noldorin" **dî instead, we must probably conclude that the final -s was maintained or re-introduced by analogy with the related word dess "young woman" (representing "Old Noldorin" ndissa). Even so, the analogy was apparently not strong enough to introduce a double ss into dîs, so the spelling **dîss is out of the question - also because it would misleadingly suggest a super-long syllable.
To avoid super-long syllables, even orthographically, we should probably also adopt the spelling with single -s whenever the plural inflection of a word makes a diphthong turn up before the double -ss. This is relevant for words with the vowel a, which in the plural becomes ai. For instance, we have argued that the word for "horn" ought to be spelt rass rather than ras, but the plural form "horns" should be spelt rais (as in the Silmarillion Appendix, entry ras). A spelling like *raiss would, at the very least, look cumbersome. The plural forms of glass, lass, nass, rhass, tass should probably likewise be spelt *glais "joys", lais "leaves" *nais "points", rhais "precipices", tais "labours". (The plural form lais "leaves" is attested in the compound Dantilais = *Dant i-lais "Fall-[of-]the-Leaves", a transient word for Autumn, PM:135.)
One could argue, then, that it would be best to adopt the spelling with single -s in all numbers and positions (I understand leading Sindarist David Salo wants to do so). It might seem to simplify some matters, but I think there are good reasons to leave double ss alone. For one thing, this spelling is clearly the dominant one in the primary sources: If we adopt this spelling as the standard, we should have to "tamper" with the spelling of only a very few words. (Indeed gas instead of *gass seems to be the only example of a word where our preferred spelling would not be directly attested in Tolkien's writings, though in a number of other cases, both variants occur - like nos vs. noss, glas vs. glass.) It should be considered a goal in itself not to edit Tolkien's material more than we really have to in order to achieve a minimum of consistency.
I would say there are also other arguments in favour of maintaining double -ss at the end of monosyllables. Everyone agrees that double ss should be retained between vowels, as in brassen "white-hot" (BARÁS); yet this adjective is derived from the noun brass "white heat", and emending the spelling of the latter word to *bras would at least slightly obscure the close relationship between the noun and the adjective. However, what I consider the best argument in favour of maintaining double -ss is, rather paradoxically, that we would then be free to simplify it to a single s - as a grammatical device of written Sindarin.
Sindarin often uses an uninflected genitive; for instance, the words aran "king" and the place-name Moria can be combined as aran Moria, "[the] king [of] Moria", as in the Gate Inscription reproduced in LotR. If we adopt a term otherwise used to describe Semitic grammar, the first word of such a construction may be said to appear in the construct state. In a phrase like aran Moria, the word aran does not simply mean "king": it appears in the construct state and rather means "[the] king [of]...", connecting with some following word. Aran is one of the words that undergo no change in the construct state, but very often, Sindarin nouns are seen to be shortened when they are so used (there are clear parallels to this in Semitic grammar). Long vowels may become short, double consonants may be simplified. The Etymologies, entry TOL2, cites the word for "island" as toll; yet it may be shortened to tol in the construct state, as is evident from the place-name Tol Morwen *"Isle of Morwen, Morwen's Isle" mentioned in the Silmarillion and elsewhere (WJ:296). Double -ss may likewise be simplified in the construct state, as is evident from the entry NÔ in Etym: Here the word for "house" in the sense of "family" is cited as noss, but immediately afterwards, Tolkien provides the example Nos Finrod "House of Finrod". Notice how noss "house" here becomes nos "house of". If we were to adopt the spelling nos throughout (citing PM:320 as a post-LotR justification for this), we would deprive ourselves of the opportunity to make these subtle distinctions in written Sindarin.
With only a few exceptions, this is indeed the system Tolkien is seen to employ. Notice the name Caradhras "Red-horn", which in LotR Appendix E is said to represent caran + rass. (Cf. also some of the other names in -ras mentioned above: Methedras, Barad Nimras.) While Tolkien cited the Sindarin descendant of Common Eldarin russâ "red-haired, copper-coloured" as ross, this statement forms part of a discussion of the name Maedros (VT41:10; the published Silmarillion has Maedhros with dh instead of d, but the sources are in agreement that ross is simplified to -ros at the end of a compound). Cf. also the final element of the place-name Cair Andros "Ship of Long-foam", translated in a footnote in LotR Appendix A, though this word *ross = "foam" or "spray" must be kept distinct from ross ?"red-haired" in Maed(h)ros. (The name Elros is translated "Star-foam"; notice that the Quenya form is Elerossë, PM:349.) We have concluded that the word for "kindred" or "family" is best spelt noss, but it is properly shortened to -nos in a compound like Drúnos "a family of the Drû-folk" (UT:385). In LotR and post-LotR sources we also find compounds incorporating loss "snow" and the related adjective gloss "dazzling white" (see VT42:18 for these independent forms), and both words then appear as -los: Mallos occurs as the name of a plant in the verse sung by Legolas in Minas Tirith ("the golden bells are shaken of mallos and alfirin"); mallos apparently means "golden-snow" or "golden-white". Uilos, "ever-snow" or "ever-white", occurs both as the Sindarin name of Taniquetil (Quenya Oiolossë) and as the name of a plant: UT:316. And while Letters:282 somewhat indecisively gives the word for "leaf" as las(s), the same source indicates that "collection of leaves" should be golas: This word is not given as *golass or even as *golas(s), for because of the prefixed element, it is not monosyllabic anymore, and in a polysyllabic word, a final -ss is regularly reduced to -s. Hence Legolas, not *Legolass.
As demonstrated above, in the Etymologies Tolkien maintained double -ss at the end of monosyllabic words in all but a very few cases. However, the orthography employed in Etym is also mostly in accordance with the rule of reducing the double consonant to -s at the end of polysyllabic words:
As far as I can see, only a very few compounds listed in the Etymologies are not in accord with this system. In the case of hervess "wife" (KHER), we find the "correct" spelling herves in the entry BES, and as for Uiloss, occurring only in the rejected entry EY, the "correct" spelling Uilos occurs in the entries GEY, OY and GOLÓS (though the first of these was also struck out). The sole entirely "wrong" example is Gochrass, the name of a sheer mountain-wall, mentioned as a transient form in the entry KHARÁS. The final element is said to be rhass "precipice", so this compound "should" have been spelt *Gochras instead. It is a relief, then, that Tolkien changed this place-name to Gochressiel! In this longer name, the double consonant ss is not final anymore, and between vowels it is rightly maintained.
As for the behaviour of words in -ss when they occur as the first element of compounds, the rule seems to be that the double ss is maintained when it is followed by a vowel, but simplified to -s before a consonant. Thus criss "cleft, cut" (Etym, entry KIRIS) survives in its full form in the place-name Crissaegrim *"Cleft-peaks", mentioned in the Silmarillion. On the other hand, "Noldorin" lhass "leaf" is reduced to lhas- in the compound lhasbelin "leaf-fading" = "autumn" (Etym, entry LAS1; in Sindarin, read l- for lh-). Therefore, though I would recommend altering the spelling of gas "hole, gap" to *gass, the spelling of a compound like gasdil "stop-gap" (GAS, DIL) should not be changed.
P.S: Of course, in this and all similar words the double ss would reappear if some ending beginning in a vowel is added, as in the collective plural *hervessath, or if we were to derive an adjective *hervessui or *hervessen "wifely": The shortening -ss > -s cannot occur unless this double consonant is absolutely final.Ardalambion Index