Implement system `bgnpcgn-isl-latn-latn-1968` (BGN/PCGN Roman-Script Spelling Convention -- Icelandic (1968)

[ronaldtse]

This issue is to implement the transliteration system of bgnpcgn-isl-latn-latn-1968.

This system is referred in the GeoNames database as isl_Latn2Latn_BGN_1968, with the system title 'BGN/PCGN Roman-Script Spelling Convention -- Icelandic (1968)'.

Tests should rely on the data extracted for the isl_Latn2Latn_BGN_1968 system in https://github.com/riboseinc/geonames-transliteration-data .

authority_id: bgnpcgn
id: 1968
language: iso-639-2:isl
source_script: Latn
destination_script: Latn
name: BGN/PCGN Icelandic 1968 Roman-Script Spelling Convention Agreement
url: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/693794/ROMAN-SCRIPT_SPELLING_CONVENTIONS.pdf
creation_date: 1968

[ronaldtse]

Spec: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/693794/ROMAN-SCRIPT_SPELLING_CONVENTIONS.pdf

[ronaldtse]

This is tricky because this system introduces conditions:

1. If the output supports those characters, edh and thorn are not converted (need to define concept of "character set")
2. If the output doesn't support edh/thorn, then transliterate them to "dh" and "th".

In the meantime, let's have this system "implemented" by inheriting the 1964 version, overriding the rules so they don't change any characters, put the second condition character changes as second rules.

I wonder if we should incorporate the concept of "supported character set".

[CAMOBAP]

@ronaldtse regarding the comment above https://github.com/interscript/interscript/issues/106#issuecomment-728642790

I wonder if we should incorporate the concept of "supported character set".

To me, it seems that we can reuse the current naming approach, but first, we need to determine what do we mean by Latn as the value of source_script and destination_script?

Initially, I thought that it should be a synonym for ASCII but in fact, it doesn't because contains characters with diacritics.

My proposal to discuss:

• bgnpcgn-isl-unic-latn-1968 - should convert to dh" and "th
• bgnpcgn-isl-unic-unic-1968 - should not convert those characters
• remove all *-Latn-* mapping to *-Unic-* if they use any other character which not included in ASCII

Let me know if it will not work

[ronaldtse]

@CAMOBAP the code Latn is ISO 15924 which includes all diacritical characters (and diacritic modifiers) and is definitely not meant to be ASCII.

This is indeed raising a good point, which is being investigated in #687 . ASCII does not currently have any mapping inside ISO 15924. Unicode does provide "Character Sets" but they are more expansive (more codes) and less specific (cannot identify an arbitrary set, such as those used by a specific language).

[ronaldtse]

Done in #694.