Re: [Indic-texts] some problems with encoding parts of akṣaras in manuscript transcriptions

Dear Andrew et al., I'm interested in how others have handled this situation. The way we handled it cataloguing the Brown, Penn, and Harvard collections of Sanskrit manuscripts was: 1. We transcribe in SLP1 which is a Romanization so allows splitting conjuncts and separating vowels from the consonants on which they depend. 2. We delete a whole vowel or consonant whenever any part of it is indicated as deleted in the ms. and add the whole replacement without comment, thus <del>o</del><add>a</add>. 3. For rendering in Devanagari or another Indic script, we thought it is not a difficult task to transpose the finer tagging of phones in a romanization to whole akzaras in an Indic script, so transposed one would get <del>ro</del><add>ra</add> [since r underscore is does not represent a Sanskrit sound SLP1 does not encode it so I'm alterning the example]. 4. For features between conuncts, <s part='I'>...g</s> <gap reason='design'> <desc>string-hole</desc> </gap>

Second, the "canonical" order of code-points and of transliteration for conjuncts with initial r has the r first. In Kannada script, however, a "flying r" is used, which occurs to the right of the other consonants in the conjunct. Sometimes there's a feature that we want to encode *between* the members of the conjunct, as in this example https://goo.gl/BqNVg9, where a string-hole intervenes between the "gg" and the "r" of "mārggaṁ". How should we encode this? I know that some of you have used "akṣarapart" to identify mātrās and other components of akṣaras, but I can't seem to get around the problem of the reversed sequence of the phonological representation and the graphic representation.