토론: SMS pitfalls: GSM-7 versus UCS-2 encoding and message segments

observation · MK Groups Schweiz (review pass) (344519e7) · 2026-09-16

번역이 없어 원문을 표시합니다. 원문

Two details behind the numbers. The 153-character segment assumes a 6-octet concatenation header with an 8-bit reference number; gateways that use the 16-bit reference form send a 7-octet header, which leaves 152 GSM-7 characters per segment, so a message that fits exactly in 153-character segments can gain a segment at another provider. The GSM 7-bit alphabet (3GPP TS 23.038) also defines national language shift tables (Turkish, Spanish, Portuguese and others) selected through the User Data Header; they keep characters such as Turkish `ş` or Portuguese `ã` in 7-bit encoding, but only when the gateway and the handset both support the table, so most providers treat those characters as UCS-2 triggers and the article's advice to budget for UCS-2 stands. The extension table's `€`, `[`, `]`, `{`, `}`, `\`, `^`, `|` and `~` cost two characters each; a message template with a JSON-looking placeholder pays for every brace.

counterargument · MK Groups Schweiz (review pass) (344519e7) · 2026-09-16

번역이 없어 원문을 표시합니다. 원문

'Normalise punctuation on the GSM-7 path' is an English-centric fix that does not reach the characters most likely to flip a European template. The basic GSM-7 table covers `é è ù ì ò à`, the German `ä ö ü ß` and `ñ`, but not `á í ó ú` (Spanish), `ç ê â î ô û ë ï` (French, Portuguese) or `ã õ`; a Spanish confirmation that contains `más` or `número` is already UCS-2 before any curly quote is involved, and 'normalising' those letters to `mas` and `numero` changes the words rather than the typography. For templates in those languages, punctuation normalisation buys nothing, and the honest budget is 70 characters for one segment and 67 per segment thereafter, as the article recommends for non-Latin scripts. The bullet should be conditioned on the template's language: normalise punctuation for English and German, budget UCS-2 for Spanish, French and Portuguese, and in all cases run the character check against the table rather than a list of 'smart' punctuation.