Dear All,

Here is a brief discussion about this issue.

By and large, we have been including all the cross-script variants in the cross-script variant analysis which (or any combination of which) could stand as a alone valid character/character sequence.

Recently IP has suggested that we may want to reconsider this where a small number of code-points are involved as that is an indicative of very small overlap between the scripts.

There are two kinds of such cases:

    1. Cross-script variant set made up of dependent characters ONLY:

    2. Cross-script variant sets which do included non-dependent characters/sequences:

Let us take a look at each of them individually:

1. Cross-script variants made up of dependent characters only:

This is the case as given in the Example 2 given by Pitinan:

Telugu ం (0C02) and Malayalam ം (0D02) are NOT variant code points. As they are combining marks and cannot form variant labels. The same applies or Telugu ః (0C03)and Malayalam ഃ (0D03)

If dependent characters (e.g. Vowel Signs, Anusvara, Visarga, Chandrabindu etc) are the ONLY cases of cross-script variants among the script involved, it is safe to assume the NONE of the labels created entirely of the cross-script variants would be valid ones. Hence we did not include them in the cross-script variants of the script pair. However, if there is even one non-dependent (e.g. Consonant, Vowel etc) character as a part of the cross-script variants, then all such cases should mandatorily be included in the cross-script variant table.

2. Cross-script variants which do included non-dependent characters/sequences:

This is the case as given in the Example 1 given by Pitinan:

Oriya ଠ (0B20) and Malayalam ഠ (0D20) are variant code points.

As both the code-points involved in this pair are non-dependent, even the smallest instance (single code-point) i.e. ଠ (Oriya ) and ഠ (Malayalam) are valid labels which look exactly alike. If we concatenate instances of same variant characters with one another, we, in theory, get infinite number labels as given below:

ଠଠ - ഠഠ

ଠଠଠ - ഠഠഠ

ଠଠଠଠ - ഠഠഠഠ

ଠଠଠଠଠ - ഠഠഠഠഠ

.....

all of these look exactly alike, belong to totally different scripts and can gain independent existence if not included in the cross-script variant set. This indicates that though seemingly the number of characters is few, it can create a large number of labels. Important thing to note here is the presence of at least one non-dependent character in the cross-script variant set.

Hence, it is proposed that:

If, in any two given scripts, all the potential cross-script variants consist of dependent (e.g. Vowel Signs, Anusvara, Visarga, Chandrabindu etc) characters ONLY, then that entire set can be ignored and no cross-script variants be proposed between those two scripts.

If, in any two given scripts, there is AT LEAST ONE non-dependent (e.g. Consonant, Vowel etc) cross-script variant character/sequence present, all the potential cross-script variants be considered and proposed between the two scripts.

Regards,

Akshat