From an internal report: "I ... found that the time separator for Arabic is ":" which should only be used for English digits IMHO (0, 1, 2, 3), but not for native numerals (٠ - ١ - ٢)
I had a look at one newspaper (http://www.ahram.org.eg/) and found they use a comma to separate times when using native numerals, which sounds like the right spearator to me."
What we have in CLDR is the following.
<pattern>zzzz h:mm:ss a</pattern>
I see two options:
1. Have alternate time patterns for different numbering systems. I see this as messy.
2. Interpret ":" in the patterns as a special symbol, like "." in numbers, that is replaced according to the number system being used for the digits. Add a new field <timeSeparator>٫</timeSeparator> to <numbers><symbols>, as below.
<timeSeparator>٫</timeSeparator> <!-- NEW -->
Not a priority for us at this time, if someone wants to pursue this, then go for it.
There's no description of the alternative idea, so reviving the proposal:
I think the main concern was that some locales use different time separators between different fields. We would just document that if they are different, timeSeparator only applies to the hours/minutes separator. The only issue we'd have would then be if a locale
a. supported different numbering systems, *and*
b. had different time separators in the same string.
This is either nonexistent or rare. (— Il meglio è l’inimico del bene —)
I am OK with the formulation in comment:17
Added the structure. TODO: add an example Generator instance
Some questions came up:
Where should it be in the LDML structure.
I looked at this, and the problem is that the separator varies with the numberSystem, not with the calendar. So while it seems a bit odd, that may be the best place for it.
How it would be used?
If (and only if) it is different than ":" in the locale/numberSystem used to format a date/time pattern, then it is substituted for : when formatting.
In other words, it functions in that case like the "," in a number pattern, which is a variable whose value is replaced by the grouping separator for the language.