the -s case

[ t h e -s c a s e ]

There have been a lot of discussions on what the -s case is used for. This case was even called the Mystery Case, because its inherent function was not known. Yet it seems that there are actually a few attested examples of the use of this case.

the -s case per se

The -s case is best known from the so called Plotz Letter (PL). This is a letter from 1966-7 which Tolkien sent to Dick Plotz; it contains declensions of nouns, a paradigm of nouns cirya ("ship") and lasse ("leaf"). The Letter is divided into three sets of cases: the first bunch, labeled (a), are obviously grammatical cases (i.e. those cases with syntactical basis), those situated in (b) are local cases (i.e. those cases expressing spatial relationship), and finally the third set, (c), contains only one case, not labeled -va case. note 1
In the Letter, nearly all of the particular cases have their names given by Tolkien except for the one ending in -s and the one ending in -n (plus the -va case on which see note 1). They share something in common: they are placed in set 2, the local cases, and they are put in parentheses below other two cases of which they seem to be shorter variants. For singular it is as follows (there is a mirror adjustment for better clearness):

All.	ciryanna
	(ciryan)
Loc.	ciryasse
	(ciryas)

From published corpus we know that the -n case is dative (see Tolkien's notes on Cirion's Oath in UT) but we do not know how to call the -s case.
As pointed out above, the placement of both of the cases suggests that they might be variants of the 'full' versions: -n of allative -nna and -s of locative -sse. Indeed according to one division of cases which combines coherency (that is spatial connexion) and orientation (direction of movement), the allative and dative belong to the same class: negative coherency with positive orientation (it means routing to something/one or activity for something/one). So both cases are theoretically used for the same thing. In Quenya however, the dative has also a syntactical role (it denotes the indirect object, cf. Sí man i yulma nin enquantuva? in Namárie) and the allative acts also as the directive (the case which expresses routing to something/one).
Now according to the same devision, the locative features positive coherency without spatial connexion, which is to say that it expresses the state of something/one being in/on something/one (cf. for example the title of Namárie Altariello Nainie Lóriendesse "Galadriel's Lament in Lorien").
Similarly as the dative has a syntactic use in Quenya, the -s case might also have some syntactic use which we, however, do not know. note 2 Whatever, if any, its grammatical function, the case in question would be primarly a locative case (sc. a variant of the locative -sse). This assumption is enforced by two little known attestations of the -s case.
Note that some some call it the short locative, this is probably the best name we can give it for the time being.

a. calma hendas

In the article The Subscript Dot: A New Tehta Usage, published in 25th issue of Vinyar Tengwar (VT), Arden R. Smith mentions a phrase written in Tengwar and published in the book J.R.R. Tolkien: Life and Legend. He transcribes the phrase as calma hendas and translates it as "a light in the eye" (with calma being "light" and henda being a variant of hen "eye" or perhaps the -a fuctions just as a connecting vowel, cf. talasse being locative of tál, VT43:16). There of course might be other intepretations of the phrase but this one seems to be the most accurate. note 3
If we assume the reading and the translation be correct, then we have a clear attestation of the use of the -s case. It is used as the locative (or at least locatively, for the case may denote possessiveness, as suggested in note 2, still falling into the same category with the locative) and it is a short variant thereof.

b. elli yas atintilar

The other example of the short locative is cited in the 28th issue of VT in the article The Bodleian Declensions by Patrick Wynne, Christopher Gilson and Carl F. Hostetter. It is a line of an early draft of Namárie taken from 'the manuscript displayed at the J.R.R. Tolkien Centenary Conference held at Keble College, Oxford, 17-24 August 1992' (ibid.; this version of the poem has not been yet published). It reads: elli yas atintilar. No translation of the phrase is given but the authors of the article concludes that it must be "wherein the stars tremble", as it corresponds clearly to yassen tintilar i eleni from published Namárie: the word eleni equates with elli (i being the definite article, not found in the elli version), tintilar with atintilar and finally yassen to yas. It is apparent that the -s case is here used as a variant of -sse, although there is one little difference: the relative pronoun yassen "wherein" is in the plural form in FM agreeing in number with tellumar "vaults" from the previous verse, whereas yas is in singular. note 4 Compare the paradigm for plural ciryar in PL:

All.	-annar
	(-ain)
Loc.	-assen
	(-ais)

Since the relavant version of Namárie has not been yet published, we cannot be sure what the preceding verse looked like. If yassen refers to plural tellumar, yas might have referred to some substantive in singular, possibly telúme or something of the kind (telúme occurs in the form telúmen in another version of Namárie (TI:284-5), the final -n may be a marker of dative/genitive singular, though it might also be a plural marker). On the other hand, Tolkien could have thought of yas (and of yasse on which see note 4) at one stage as of not agreeing in number, which is not surprising. Note for example that the -va mentioned above does not seem to have any plural form. Yet Tolkien claims the contrary in WJ:407. There are some instances of adjectives not agreeing with their governing substantives (see VT39:17).

the -s and -r case; locative and dative

The fate of final -s was not always the same in Tolkien's Quenya, he seems to have been changing his mind to and fro. Compare these two words: olor and olos, both meaning "dream". The former is from Etym (it also occurs in QL), and based on other examples therefrom, it can be postulated that final -s became -r (via -z) in Quenya at that time. The same conclusion can be abstracted from Tolkien's notes published in Parma Eldalamberon #12 p. 19, 20: 'It is to be noticed that -þ- > s did not suffer medical voicing [sc. s > z > r], as the other s'. There are (at least) two sources of the s: one from the original s and one from the þ (th), their development differed a little bit. While the original s was sonorized, the þ (th) remained medially. According to QL both were sonorized finally, the entry Tulkatho > Tulkas (LR:395 s.v. TULUK-) suggests that in the Etym era, final þ (th) did not change after having become s. Now the word olos from Unfinished Tales p. 4 ch. II suggests that the final position was not strong enough to trigger voicing in Tolkien's later vision (for more on this see Helge Fauskanger's The Evolution from Primitive Elvish to Quenya).
Given this piece of information, the case ending which we know as -s might have sometimes undergone rhotacism and taken the form -r. There are instances of the case ending -r that might be a rhotacized form of the short locative -s, though sometimes it seems it may be of different origin.

a. the en- declension

Aside from the Plotz Letter there are two different published charts showing Q(u)enya declension system. The first one is the so called Entu, Ensi, Enta Declension (ED), written most likely between 1924 and 1936. The forms for dative and locative are as follows (particular cases are not named in the manuscript but it is not very probable it would be otherwise) note 5:

[dative]	entur
[allative]	entonta
[locative]	entosse, (-os)
	entunye

We can see here again the -s form arranged as a variant of the locative. Again, it is given in parentheses (note that the dative is not, perhaps because the dative ending here is not etymologically related to the allative here).

b. Oilima Markirya 1

The first version of the Markirya poem called Oilima Markirya (OM1) as published in MC:220-1 contains a few following lines which have something to do with the locative case:

(1)	veasse lúnelinqe
	upon the blue streams of the sea
(2)	kaire laiqa'ondoisen kirya
	the white ship lay upon the rock
(3)	ailissen oilimaisen
	upon the last beaches
(4)	alkarissen oilimain
	in the last ray of light
(5)	ailinissen alkarain
	upon the shining shore

It is apparent from these lines that the suffix for the (full) locative was -sse in singular and -ssen in plural, which agrees with PL. The form -(i)sen in ondoisen (2) and oilimaisen (3) seems to be an alternative locative plural suffix, presumably the plural form of the short locative -s (not found here; or of *-se or -r). It did not undergo rhotacism, which is interesting. Perhaps due to the analogy or maybe because of the stress (cf. VT27:37-40). However, it cannot be said that there was no rhotacism rule in Quenya at that time. The -re in kaire (2) is most likely the feminine pronominal suffix, a rhotacized form of -se (see LR:385, s.v. S-); the initial element kai- is no doubt a bare form of the stem KAY(A) (QL:46L, LR:363), here used as a preterite. The suffix -re refers to kirya, cf. also kirya kalliére "the ship shone" (OM1).
In light of this, it is possible that the short locative ending -s is derived from original *-th (-þ) instead of *-s, because medial -th (-þ) did not suffer medial voicing as noted above; the full locative -sse may show doubling of this element.

c. Oilima Markirya 2

Another version of the Markirya poem, also called Oilima Markirya, (OM2) as published in MC:213-4 contains these lines:

(6)	níve qímari ringa ambar
	the pale phantoms in her cold bosom
(7)	lúnelinqe vear
	in the flowing sea
(8)	silda-ránar
	in the moon gleaming
(9)	minga-ránar
	in the moon waning
(10)	lanta-ránar
	in the moon falling
(11)	laiqa ondolissen
	on the green rocks
(12)	píke assari silde
	on bones gleaming
(13)	óresse oilima
	in the last morning

From the examples óresse (13) and ondolissen (11), it is clear that the (full) locative remained unchanged (in form, its origin may be different, see below), sc. -sse in singular and -(i)ssen in plural (note that laiqa'ondoisen was changed to laiqa ondolissen).
The problem lies, however, in the ending -r (examples 6-10) or which -ri in assari (12) may be a plural. There appeared (at least) two interpretations: (a) it is an ending of the dative (on this see VT36:21-2) and it is here used locatively, or (b) it is etymologically an ending of the locative which was used as the dative in yar "to whom" (Nie; on this see VT28:19-21). The solution may actually lie somewhere in between.
It will be useful to write out here relevant lines from the final version of Markirya, (FM; on translation see Helge Fauskanger's analysis thereof):

(14)	ringa súmaryasse
	in her cold bosom
(15)	ëar-kelumessen
	in the flowing sea
(16)	isilme ilkalasse
	in the moon gleaming
(17)	isilme píkalasse
	in the moon waning
(18)	isilme lantalasse
	in the moon falling
(19)	ondolisse morne
	on the black rocks
(20)	axor ilkalannar
	on bones gleaming
(21)	métim' auresse?
	in the last morning?

As is apparent, the ending -r from OM2 was always replaced by -sse in FM (compare lines 6, 8-10 with 14, 16-18), except for ear-kelumessen where we see the plural ending -ssen (see lines 7 and 15), which may be a mere adjustment accompanying the changes introduced (such as different words used). Therefore it may be theorized that the -r is a locative ending, presumptively a rhotacized form of -s. However, if we compare lines 12 and 20, we can see that the plural form -ri in píke assari silde was substituted by -nnar, an allative plural ending, in axor ilkalannar. note 6 This fact would speak for the theory which assumes that the -r is the dative in origin, used locatively sometimes. It could be also possible that there are two distinct cases, short locative and dative, with the same phonetic realization, that is to say that the short locative ending -s might have undergone rhotacism and merged with (original?) dative ending -r. They might have differed in plural, though: dative pl. -ri, locative pl. *-ren or -sen (without medial voicing as in OM1) or -ssen (cf. the words nis and nisse both meaning "woman" and both having plural nissi, LR:377).
It is difficult to find a way out of this. For example, if we assumed that the endings -r and -ri are endings of the locative, we would have to answer the question what did the dative look like? The ending -n which we know from PL might be a possibility but it was generally used for something else in this period (for genitive in ED, but see VT36:22 on enden, for ?subjective in BD, see VT28:13-7), yet this ending is possible, in The Lost Road, the ending -n is used for both genitive and dative: ilúvatáren "of-Ilúvatar", nin "for me" (LR:47, LR:72).
The ambiguousness might be dealt with otherwise. It may that the use of the use of -ri (12) was just Tolkien's 'mistake' or Qenya of the time used the locative in such a case, that is to say where the dative/allative would be expected instead. And hence the change to -nnar in FM (20) could have been probably only a correction. If we take a look on the whole phrase which reads: Man tiruva [...] anar púrëa tihta / axor ilkalannar which means literally "who shall watch a smeared sun blink[ing] towards bones gleaming?", the allative denotes here the direction of the blinking of the sun, while the (supposed) locative in OM2 would rather suggest that the blinking sun was situated upon the gleaming bones.
Here something must be said on ambar (6). I have so far assumed silently that it is *amba "bosom" and -r a dative/locative singular ending. That there is no apparent "her" in the word (unlike in case of súmaryasse "in her bosom": -rya) was clear enough (though of course the morpheme denoting "her" might be present, having fused with the morpheme with the suffix -r being its realization). The word *amba is asterisked because it is not attested in any published material. The word is not, however, d'accord with the Qenya Lexicon which has Ambar "breast" (QL:30L; the note in parentheses following the word is also interesting, it reads '-s or -r'!). Now there are several possibilities: (a) the word *amba is a variant of ambar and the -r is indeed an ending; (b) the word ambar "in her bosom" exhibits contamination of the case ending -r (dative/locative) with the word ambar "breast"; (c) the word ambar is in nominative which may be used also as inessive.
We are told that the early Qenya and Goldogrin (sc. Gnomish) had three noun cases: inessive or nominative, genitive, and allative or dative (GL:9). This means that nominative was 'occasionally used by itself as a locative' (GL:9). note 7 Since the nominative singular stem is the bare, uninflected, form of the word, the same holds for inessive. In his analysis of the poem Narqelion, Christopher Gilson notes that such an inessive function of nominative may be seen in V'ematte sinqi Eldamar *"as they call to mind the gems in Eldamar" (VT40:23-4), the word Eldamar does not show any inflexion, hence it might be nominative and inessive (unless final -mar fused with -r). (Another, later, inessive use of the nominative may be seen in Aia María: sí ar lúmesse ya firuvamme "now and at the hour in which we will die" (VT43:34), although the lack of any locative suffix (or preposition) at ya might be due to its presence in lúmesse.)
Now the same situation may be witnessed in the phrase níve qímari ringa ambar: although Q(u)enya of the time distinguishes nominative (bare or (?) -n stems) from locative (-sse or (?) -r stems), it is possible that the nominative could have inessive function in certain constructions (or in verse when needed). note 8

d. Nieninque

Another poem mentioned by Tolkien in his essay A Secret Vice is Nieninque (Nie). Based on the use of tande "thither", it seems to be written after OM2 (together with the poem Earendel which does not seem to be important for this discussion). It contains these lines:

(22)	elle tande Nielikkilis
	thither came little Niéie
(23)	yar i vilya anta miqilis
	to whom the air gives kisses

There is no full locative form in this poem but it was probably -sse again, this ending does not seem to change. The form yar "to whom" is clearly the dative of ya "which" (see VT43:34), the same dative ending is given in ED. The form tande may be analyzed as ta "that, it" (LR:389, s.v. TA-) and -nde, which is just a allative ending or a form of -nda (on which see below).
The morphs -r and -nde might be of the same origin. In Etym, we can find preposition mir "into" (LR:373, s.v. MI-) and pronoun (?) tar "thither", said to be derived from tad (LR:389, s.v. TA-). Aside from this, there is adverb öar "away" which 'shows addition of the ending -d (prehistoric -da) indicating motion to or towards a point' (WJ:366). The ending -d seems to be the oldest allative suffix; the preposition mir may contain it as well (besides mir there is also minna of the same meaning). The suffix -nde might show a nasal infixed form of -d.
Whether the -r cases in ED and OM2 are of the same origin cannot be said. Tolkien is known to have come up with wholly different etymologies of the same thing, so the -r of ED and OM2 might be from something else. It must be pointed out that the allative ending in both ED and OM2 is -nta. Devoicing of original -d is not very likely, but on the other hand, the allative and dative endings of the time need not be related. However, as was already shown, the ending -r might be a rhotacized form of short locative -s in OM2 and there is a possibility that at some stage of external development of Quenya the ending -s was used for the dative.

e. the Bodleian declensions

The other one of those two charts mentioned above is the so called Bodleian Declensions (BD). Being probably from 1936, it is therefore later that ED and Nie. Particular cases are not again named. The forms relavant for this discussion are:

	[singular]	[plural]
?	kiryas	kiryais (aisi)
[locative]	kiryasse	(-sser) ·n
[allative]	kiryanda	(-andar)

Analysts of the paradigm call the -s case short locative, based on PL and the use of yas in a draft of Namárie (see above). They take the -r case in lines 6-10 to be a rhotacized form thereof. The short locative, according to them, has a syntactical role as the dative and they compare it with the dative case of the early Adunaic case system. The Adunaic dative singular is -s, plural -sim.
The ultimate origin of the ending probably lie in Greek dative plural si (Attic -ais, Ionic -aisi, identical to Quenya forms!). It may be very interesting and important to know that the -s- in Greek dative plural (it also appears in Latin) is rather typical for locatives in other Indo-European languages, cf. Sanskrit loc. pl. -su. In Greek, the dative is used also as the locative. However, unlike Quenya, Greek does not have a distinct locative case. Tolkien might have been inspired by the Greek situation and Q(u)enya dative -r might have been the locative in origin and sometimes even used in this original role when needed and therefore the -s here in BD would be an un-rhotacized form.
In the chart, there is no short locative per se. In other charts (PL, ED) we have a short variant of the locative grouped with the full locative. It is not perfectly certain that the -s ending is a short variant, as was suggested in VT28. The dative, sc. the -s, could be of another origin, perhaps from *-th because it did not undergo rhotacism, even not intervocalically (-isi). Note that the same origin was offered to -sen from OM1. Taken from other side, it is appearent that certain s's are not intervocalically sonorized and then rhotacized in QL.
As suggested in the c section above, it is possible that the dative suffix contained the element r from the beginning and the locative suffix contained the element s or rather th (þ). In the singular, the dative ending might have been -r and the locative ending -r, having undergone rhotacism, or -s when Tolkien decided for non-rhotacism in final position. In plural, it would be -ri for dative and -sen for locative. The -s dative in BD may indeed be locative in origin, being inspired by Greek. Compare also Gnomish dative/allative ending -r which is compared with Qenya -r adverbial (and with -l in nouns; QL:10) and ar "at, to" (PM13:110) which would speak for the dative in -r (and Tolkien might have come up with different origin of this ending, being from -d).

sê, jê, men

The publication of Vinyar Tengwar 43 shed more light on the locative problem. The issue presents various versions of Pater Noster (PN) and Ave Maria (AV) translated to Quenya by Tolkien in the article "Words of Joy" by Patrick Wynne, Arden R. Smith and Carl F. Hostetter. Various versions feature various ways of expressing "in/on something/one":

menelle 'in heaven': menelle (At. I), menelze (At. IIa), and menelde (At. IIb-V) are all locative case forms of menel 'heaven', each exhibiting a different phonetic result of contact between the final consonant of menel and the locative suffix -(s)se: *menel-se > menelle, menelze, menelde. Compare the locative forms of cemen 'earth' in this same line: *kemen-se > cemesse, cemenze, cemende [...]. The adj. menelessea (At. I deletion) appears to be derived from loc. menelesse, in which the locative suffix is preceded by an epenthetic or stem vowel; the coexistence of such forms side-by-side with directly suffixed forms such as menelze, menelde is demonstrated by an unpublished declension of tál, c. 1967, which gives the locative forms as talasse and talse. (p. 16, At. I etc. being different versions of Quenya Pater Noster)

[...] olesse 'with you' can be explained as the preposition ó- 'with' (in reduced form in unstressed environment) + (shortened) 2 sg. ending -le 'you' + locative sg. ending -sse 'in, at'. The subsequent forms carelye and aselye are presumably to be explained similarly as variant prepositions (as-, car-) expressing accompaniment, with the (full) 2 sg. ending -lye, but without the locative ending. (p. 29)

The preceding versions [of olesse] le se and lese give separately and then combine the independent form le 'you' and se 'at, in' (the root of the latter is given [...] as sê 'at'; cf. the locative ending -sse). (p. 30)

[...] the same entry sê 'at' [...] also gives men as an alternative gloss. The gloss men was subsequently struck through [note on p. 36:] [a]nd replaced with jê; men being subsequently reassigned as a gloss from 'with (instr[umental])' (p. 35)

It is obvious that the locative ending was originally a postposition (so was the allative/dative and genitive) derived from the root sê. This postposition when added to a consonant-ending word was partially assimilated (voiced) if preceding consonant was voiced (menelze) or wholly assimilated (menelle). The z-forms have to be in archaic/Book Quenya because Quenya of the Third Age did not have the phoneme z while the geminated forms have to be Exilic Quenya.
The variants cemende and menelde are different. The authors of the cited analysis take them to be phonetic results of contanct between sê and sonors n and l.
I think it may be hard to explain how s could become d if from original s, sc. from root sê. However, it may be easily explained if the source of locative preposition se would be th (þ) as I suggested above. This th (þ) if in contact with preceding l would result in ld and perhaps even in nd if in contact with n (cf. QL:19 where Tolkien states that liquids + þ produce ld and rd, although nasals + þ produce nt -- Tolkien might have later decided that all sonors trigger voicing, therefore nasals + þ > nd). That may also explain why lese was allowed to stand. On the other hand, Q(u)enya might have been in a stage when no medial s of whatever origin changed, cf. nísi "women" (p. 31) and and base NIS- (LR:378). note 9 Also, the root of the locative se is explicitly given as sê not *þê. But cf. this statement: '[...] the notes on the verso of At. V attribute úsahtie to a different stem: "saka- "draw, pull'; þ/sahta 'induce': úsahtie 'inducement to do wrong'"' (p. 23). How (inter)dental þ could arise from alveolar s in Quenya is somewhat unclear, one would expect the reverse development which would mean that the root was *þaka- instead (did Tolkien or Eldar confuse it with the root thag- "oppress, crush, press" mentioned on p. 22?). It may therefore be possible that at one stage se was derived from þê in Tolkien's mind. It must be also noted that talse was a permitted form which did not undergo any change to *talle or *talze but even not to *talde which speaks for the proposed theory though even against it. Possibly the de in cemende was of different origin that the one suggested (allative -da might be possible, cf. tande in Nie, though it does not do fit well semantically).
The authors of the cited article attribute the preposition as "with" to the conjuction ar having 'a Common Eldarin form as 'and'' (p. 30). 'In the current Quenya example, as- 'with' may be a prepositional reflex of C.E. as and related to the Quenya conjuction ar 'and'' (ibid.). Though this may be possible, one may wonder why the CE form as did not become Quenya as, either because CE s did not change finally or in every position (cf. nísi) or at least in the versions At. I and At. IIa wherein hyáze occurred being derived from the root AS- "warmth" (see p. 18). Since the note on CE as is from c. 1968, it is more probably that ar was still a derivative of the root AR (LR:349) at that time.
It may be possible that the preposition as- is related to the base sê 'at, in' rather that to the base as- "and" or perhaps aselye is a *"with" + (short) locative -s, compare the alternative olesse being compounded of o- "with", le "you" and locative -sse. The locative expresses the state of being in/on something/one, maybe it likewise expresses the state of being with something/one. note 10
The authors of the article mention two elements: jê and men both connected with the locative. The former may explain the peculiar locative forms as entunye mentioned above (ED), the latter may be the origin of Adunaic men which was replaced by dative si (SD:311), which may be locative in origin, as argued above. And speaking of Adunaic, this language has a preposition zê "at" which seems to Quendian in origin, being probably from the same base sê as Quenya se. The initial z would rather seem to be from original s than from th (þ), though we do not know the fate of the Elvish words beginning with þ borrowed by Men.

From the angle I look at the problem of the case endings -r and -s, far too many points remain uncertain. There are evidences speaking for some theories but also evidences agaist the same proposed theories. Hopefully the future publications will further our understanding and knowledge of Tolkien's invented languages.

notes

[ note 1 ] The -va case is not named in PL but it is undoubtedly the possessive-adjectival case whose origin and usage was finally clarified by the publication of The War of the Jewels, see pages 368-9.

[ note 2 ] There have been many theories as for what the -s case is used and what is its name.
One of the names and theories was Respective; this name was usually used when referring to the case. To quote from Anthony Appleyard's Quenya Grammar Re-examined (some editing intoduced):

The role of the case that I call 'respective' may relate to Locative as Dative relates to Allative: when a noun A is Allative, something goes to(wards) A, and when A is Dative, some object or action or favour 'goes to' A semi-literally or figuratively. Perhaps -s means "concerning", "with respect to", "paying attention to", and (most often) "about" in "think or talk or write about", whence my name for it: when A is Locative the action is at A but does not move towards A, and when A is Respective the action concerns but does not directly affect or 'go towards'.

Another theory was that the -s case denoted verb "to have", sc. it expressed possession, possessiveness. Quoting from Ryszard Derdzinski's Elfling message, 19 Jan 2001 (some editing introduced):

Ownership or possession expressed by the verb have in English, in Finnish is expressed by using verb corresponding to English be and putting the logical subject into a case ending with -lla or -llä; thus, 'I have a dog' is in Finnish Minulla on koira ("at me (there) is (a) dog").
As I stated long ago, in my opinion the same function has the Mystery Case that can be called Adessive. Because the Mystery Case belongs to the 'Location' sub-group of the Plotz case system it is possible it has similar meaning as the Finnish 'Adessiivi' -lla. If so, late Quenya 'I have a dog' would be Nis ná ronya 'At me is a dog' and not Haryan ronya 'I possess a dog'.

There might be something in favor of the latter theory: Tolkien liked Finnish and was influenced and inspired by it. It therefore might be possible that the -s case was used this very way, though, as Ryszard noted, there is the verb harya "to possess" (LR:360, s.v. 3AR-). Note that this verb occurs in the so called Merin Sentence: haryalye alasse "you have happiness" (the validity of the sentence is not wholly clear). The -s structure was not used in it, which does not of course prove it was not possible.

[ note 3 ] Arden R. Smith notes that it is curious that the word calma is written with the tengwa no. 3 which is normally used for qu. An alternative reading with qualma instead of calma might bring a new interpretation, but if related to the stem KWAL "die in pain", it is hard to offer a satisfying explanation.

[ note 4 ] The word-order of the line also differs: elli yas atintilar literally "stars wherein tremble". Most likely it ought not to have any influence on the interpretation of the phrase and the word-order is only accomodated for a poetic style (cf. published Namárie in LotR with prose Namárie from RGEO). Since this elli-version of the poem has not been yet published, it is hard to say what metric pattern this version had.
The plural elli is not found elsewhere, the published text has (i) eleni. In WJ:362, it is said that a poetic plural of elen (besides 'normal' plural eleni) is eldi. It is possible that this word became later elli by assimilation (cf. nulda and nulla "secret", both from ndulna, see LR:355, s.v. DUL-). Alternatively, Tolkien might have imagined a slightly different-in-form word for "star": *elle (final vowel is uncertain) or something of this kind. The word atintilar is almost identical with tintilar from the published text. As for the a- prefix, it is hard to judge its role here. There are several a- prefixes in Q(u)enya: (a) an intensive prefix, equal to sa- (see QL:81L, s.v. SA-); (b) a prefix denoting completeness (see LR:390, s.v. TALÁT); and a prefix used for infinitives "if the noun is the object not the subject" (see MC:223); (d) besides it might be a prefix for subject/object "it" (it is not attested as a prefix but as a unitary word, see the phrase En a túvien from SD:57). All in all, the most likely seems to be the first possibility, hence atintilar may mean *"greatly tremble".
The authors of the article The Bodleian Declensions also note that the phrase elli yas atintilar "was struck out, and two other versions of this line appear on the manuscript, both allowed to stand: yassen elli atintillinar and yasse tintilar i{n} eleni (VT28:11, note 5). "The latter phrase almost achieves the form of the published text, save for sg. yasse instead of pl. yassen" (ibid.). The word-order was changed, the short locative replaced by 'full' locative (it remains questionable whether this was done for metrical reasons or the -s case was not allowed to stand here), the form atintillinar is somewhat curious. It is noteworthy that Tolkien was toying here with the idea that the definite article take the form in in plural and/or before a word beginning with a vowel. The practice of attaching an -n to particles used antevocalically was common in Qenya Lexicon, cf. Kuluvai ya karnevalinar (Narq., VT40:8) with Eldi yan Indi (QL:43L) and ya(n) "and" (QL:104R). The antevocalic/plural form in is also seen in a very late source in the sentence Mana i·coimas in·Eldaron?' maquentë Elendil (PM:403).

[ note 5 ] The locative case ending -nye does not appear anywhere else. -- In front of entunye there is (in the entu paradigm, it is similar in the others) struck out entos and the nye form is underlined. The short variants of the locative were not rejected, though, see the -os in parentheses.

[ note 6 ] The fact that the case ending was added to another respective word than in the OM2 does not most likely play any role. The same divergence is seen in lines 16-18 wherein the case ending was also added to different words that in OM2 (in all cases it is the participle). For the discussion on this phenomenon see Helge Fauskanger's Quenya Course, l. 17.

[ note 7 ] The inessive is the locative, it is another name for it. The Finnish case which expresses location within something/one is called inessive. It cannot be a mere coincidence that its suffix is -ssa/ssä.

[ note 8 ] Theoretically, all the -r lines could be explained this way.
The word vear in line 6 would have to be in plural form. There is some probability of this, despite the singular translation: cf. line 1 of OM1 veasse lúnelinqe, corresponding to line 6, which is translated as "upon the blue streams of the sea" ("streams" in plural!) and ëar-kelumessen of FM, also corresponding to (6; "streams" again in plural!). Although the word vea means "sea", it is possible that here in the poem it might mean "seas" (sc. vear).
In lines 8-10, the word *ránar would have to be a variant of Rána "the Wanderer" (SA) with the suffix -r being an agental ending.
Finally, the assari which was treated as if containing locative plural -ri. Likewise, I assumed silently it was so; QL lists As "bone" with stem ass- (p. 33L). Hence the second a in assari would have to be an add-in vowel for connecting the stem with the case ending. As an alternative we could suppose existence of *assa being a variant of as. But we could imagine another variant of the word: *assar, and then assari would be nominative (and hence inessive) plural. Also, the whole -ri might be a plural ending (cf. VT27:23 for similarly formed dual aldaru).
Having this, we would easily avoid all the ambiguousness of the ending -r. However, these interpretations are probably too overcombined and not very probable. In fact, it is not even certain that the word ambar is really nominative in its inessive function, cf. Gn. am and N. amb (struck out; PE13:173).

[ note 9 ] Though this may be explained otherwise. nísi may be featuring the feminine ending *-the (-þe), its masculine counterpart may be seen in Tulkatho derived from the stem TULUK- (LR:395). This may be related to Adunaic -th 'often met in feminines' (SD:427). Cf. also Westron arantha/aranthe "queen" mentioned in TyTy17:37.

[ note 10 ] Ryszard Derdzinski wrote on Elfling, 3 Nov 2000 (edited):

Many have also tried to guess how to say in Quenya 'X with Y'. The 'with' preposition has been expressed in many Quenya compositions as arwa (Ety 360; 'in control of; posses[s]ing') or unattested *ho 'together with'. For example the sentence "With Manwë dwells Varda" (S 28) might be translated as *Arwa Manwëo Varda marë or *Ho Manwë Varda marë.
In my opinion (remember - it is a mere hypothesis and I know it can be completely wrong) in such phrases the so called 'Mystery Case' should be employed. Because it is supposed to bring the 'locality' meaning (as Dative expresses a kind of movement towards expressed in full by Allative), it may be its function to describe state of being 'with Y'.
So, in my opinion "With Manwë dwells Varda" could be translated as "Manwes Varda marë". Another example could be phrase from "Ava Maria" prayer, translated by me into Quenya long time ago, where we have "Lord is with you". In light of my hypothesis it could be translated as "Heru [ná] elyes"[.]

sources used

Appleyard, Anthony:

Quenya Grammar Reexamined, TolkLang site, <http://www.dcs.ed.ac.uk/misc/local/TolkLang/articles/Appleyard.Quenya>

Derdzinski, Ryszard:

Finnish adessiivi and Quenya Mystery Case, Elfling mailing list, 2001

Quenya 'Short' Locative, Elfling mailing list, 2000

Erhart, Adolf:

Indoevropské jazyky, Srovnávací fonologie a morfologie, 1982

Fauskanger, Helge K.:

Quenya Course, Ardalambion, <http://www.uib.no/People/hnohf/qcourse.htm>

The Evolution from Primitive Elvish to Quenya, Ardalambion, <http://www.uib.no/People/hnohf/qevolution.rtf>

The Markirya Poem, Ardalambion, <http://www.uib.no/People/hnohf/markirya.htm>

Mäkinen, Panu: