- The following message first appeared on Elfling; it reappears here at Mns's request, with several errors now corrected (most of them pointed out by Mns in his response to the message).
None of Tolkien's published writings say anything about the system of writing used to represent the language of the Teleri. Lacking such evidence, one can do little more than speculate; but not, I think, beyond the bounds of probability.
There can be little doubt but that the language of the Teleri was reduced to writing in Eldamar, if not by the Teleri themselves, then by the linguists of the Noldor whose study had expanded to include that language; the only language in Aman, other than the little-known Valarin, to show significant differences from the Noldorin Quenya. Indeed, the preservation of such scraps of Telerin as survive would have to be attributed to such 'loremasters' who came with the exiled Noldor to Beleriand, and transmitted their knowledge to the Elves and Men of Middle-earth.
Either way, the writing of Telerin would doubtless have been primarily based on the tengwar writing system devised for Quenya. Where it differed, it would be either because Telerin possessed different sounds from Quenya, or because a more consistent mode of representation of the sounds of Telerin could be achieved by a different organization of the tengwar -- though admittedly this would be chiefly of interest to the 'loremasters'.
In addition to the standard usages of written Quenya (Parmaquesta), those who devised a system for writing Telerin would also have had access to the more strictly phonetic writing-systems of the Elvish linguists, which were (at least optionally) fully alphabetic (cf. Vinyar Tengwar #39, p. 8). One such alphabet (not necessarily the only one) is shown in the first version of Tengwesta Qenderinwa (Parma Eldalamberon #18, pages 30 and 32). Something of the nature of these variant alphabets can also be deduced from the tengwa-alphabets devised for the representation of Sindarin, especially the 'mode of Beleriand', which presumably dates back to the First Age; and the Sindarin usage, being closer in certain respects to Telerin, can suggest possible values for Telerin tengwar.
The signs required for the representation of Telerin are:
Consonants: p, b, f, m, v (w); t, d, T, n, r, l, s; c (k), g; y, h. Vowels (if represented alphabetically) a, e, i, o, u. Diphthongs: signs for "following i" and "following u".
Of these, Quenya provides unambiguous signs for p, f, m, t, T, n, l, s, c, and h: the tengwar parma, formen, malta, tinco, Tu:le, nu:men, lambe, silme, calma, and hyarmen. These tengwar agree with those of the 'mode of Beleriand' except for m and n, for which the Sindarin mode used vala and o:re; and that because -mm-, -nn- are very common in Sindarin (arising not only from *-mm-, *-nn-, but also from *-mb-, *-mp-, *-nd-, *-nt-), and so it was convenient to use the tengwar with double lu:va (malta, nu:men) to represent mm, nn.
This usage would not be necessary or even particularly convenient for Telerin, where -mm-, -nn- were less common, and so I judge it likely that Telerin agreed with Quenya usage in this respect.
Telerin could not, however, agree with Quenya with respect to the representation of d, b, g, for which Quenya had no separate letters, as they never appeared except after nasals, l, or r. In Telerin they were much more common, and often appeared initially. I judge it highly likely that Telerin would have used the letters ando, umbar, anga for these sounds, just as in Sindarin (and in the TQ1 alphabet); and consequently would have represented -nd-, -mb-, -g- by means of the nasal 'tilde' over the same three letters.
That leaves doubt only as to the representation of v, r, and y among the consonants. R could be represented by either o:re or ro:men. Some varieties of tengwar use the one, some the other, and in some modes the value of r is split between the two letters on various principles, e.g. o:re used only at the end of a word, or only at the end of a syllable. This last distribution is found in Tolkien's transcript of the Quenya "Nama:rie" (RGEO p. 57), where I suspect the distribution really reflects the positions in which r is deleted in British English; there are no grounds, in any of the descriptions of Quenya phonetics I know of, for supposing a phonetic distinction related to this environment. The two r's could also be split on etymological grounds, e.g. distinguishing r < *r, r<*z, and r<*d (the last of which does not occur in Telerin). I imagine such a distinction would appeal to the loremasters! TQ1 uses ro:men to represent r < *z (in which case it would be better named a:re), and o:re to represent r<*r.
The phonetic equivalent for Telerin v in Quenya would be represented by the tengwa wilya (originally [w]). But I think that the tengwa vala would be a better choice, for two reasons: first, because it would 'explain' (on grounds other than those of imitating the orthographic style of written Latin since c. 1590) why the sound is represented in transcribed Telerin by "v"; and also because it allows the consonant to fit in with the labial series p, b, f, m, rather than using a shape proper to the labio-velar series (even if [w] is, strictly speaking, a labio-velar sound). With v=vala, Telerin could be written without any consonants of te:ma IV at all.
As for y, Quenya usage is the letter anna (representing originally the lost consonant '3') with two dots underneath, thereby relating it to the tyelpete:ma. But in Telerin, which had no tyelpete:ma, the two dots would seem superfluous. On the other hand, using anna alone creates a visually misleading association with the calmate:ma. The best solution, I think, would be to use an entirely different letter, namely yanta -- which also represents [j] in TQ1 p. 30. In fact, the resulting consonantal alphabet is almost identical to that shown in TQ1, except that it lacks signs for independent  (TQ1 noldo), [x] (TQ1 harma),  (TQ1 anna), and labio-velar [w] (TQ1 vilya) distinct from labial v, and uses hyarmen for [h] instead of halla.
In a mode with vowels represented by tehtar, exactly the same usage as Quenya can be employed; long carrier for long vowels, tehta+yanta for diphthongs ending in -i, tehta+u:re for diphthongs ending in -u. Other markers and special characters -- wavy line above the letter for a preceding nasal, line underneath the letter to show doubling -- could also be used. The letter esse to denote a double s would probably be in use; alda to denote -ld- would probably not be (as the sound was lost except in cases of secondary syncope) but it might be pressed into use to represent the frequent -ll- (as in some of the English tengwar modes).
Telerin has so few consonants, however, that it could easily and profitably be written with a fully alphabetic script. The best candidates for vowel-letters are, I think, those used for Sindarin; but their use requires a slight change to the consonantal system.
The relevant Sindarin vowels are a short carrier for "i" (often, but not always with a dot); yanta for "e"; u:re for "u"; anna for "o". For "a" the letter "osse" (a shape like the letter c) was used. I think this shape represented the lu:va of the letter anna without the telco. (Similar telco-less forms of tyelle 6 are found elsewhere; a telco-less o:re (which, for the sake of convenience, until a more suitable name appears, I should like to call "pero:re", i.e. "half-o:re"), shaped like a backwards c, is found in the TQ1 alphabet (p. 32) representing the sound of "o"; a half-vala ("pervala") is used in some phonetic tengwar transcriptions of English to represent "w"; and a half-vilya ("pervilya") is used in the same transcriptions to represent the vowel sound of schwa.)
If yanta is used for "e", then a different sign must be found for [j]; the form found in some alphabetic Sindarin scripts is a long carrier (i.e., almost like the letter j, though sometimes without the dot), and this seems to be apt. Diphthongs are shown by the vowel-letters with either the diacritic of two dots for following -i or an elaborate curl (something like a hyperactive tilde) representing following -u. Long vowels are indicated by the !andatecta "long mark".
We are told that even those who did not speak Quenya used the Quenya names for the letters. This must have posed a problem for the Teleri -- whose nearest approach to the word "Quenya" itself could only have been "Cuenia", turning two syllables into four! But if the Quenya names are given Telerin forms, suitable names can be found for almost all the necessary letters. Imaginary possibilities are shown below, separated below by telli (Q. tyeller).
Telle 1: 1. t: tinco, as in Quenya (or telpe "silver") 2. p: parma, as in Quenya (or pesse "feather") 3. c: calma, as in Quenya
Telle 2: 5. d: du:men "west" 6. b: bala "Vala" 7. g: galada "tree", gampa "hook" or garda "region" (golodo "Noldo" would probably not have been used, for obvious reasons)
Telle 3: 9. T: Tu:le, as in Quenya, or Tarmen "south" 10. f: formen, as in Quenya
Telle 4: not used
Telle 5: 17. n: ne:n "water" (no suitable name among the Quenya letter-names) 18. m: malata "gold"
Telle 6: 21. r(<*r): o:re "heart" 22. v: vilia "air"
25. r(<*z): a:re "sunlight" 27. l: lambe "tongue" 29. s: silime "starlight" 33. h: halla "tall" or harma "treasure". 35. y, -i in diphthongs: ianta "bridge". 36. -u in diphthongs: u:re "heat".
A more thoroughgoing Telerinization of the letter-names could be made in order to fit the circumstances. lifestyle, and preferences of the Teleri, especially considering the "large vocabulary of sea-words" in Telerin (The Peoples of Middle-earth, p. 365). For instance, ciria "ship" might be substituted for calma; vola "long wave" for vilia; or londe "haven" (or linda "one of the Teleri") for lambe.
Due to the close resemblance in sounds between Telerin and Latin, this theoretical Telerin alphabet could easily be adapted for the representation of Latin; the only additional provision required would be a representation of the sounds [kw], [gw], and this could be easily enough achieved by using the letters quesse and ungwe from the quessete:ma. In fact, an almost exact imitation of the pleonastic spelling "qu" in Latin could be achieved by combining the letter quesse with a w-curl.