Peter Kirk continued: > I did read it, but it didn't deal with the issue I was concerned about, > of multiple combining marks. And I was concerned about that issue > because that was the major concern expressed in the earlier discussion > on variation selectors, and presented as the decisive reason why > variation selectors cannot be used with combining marks.
And I agree that that is the (or at least "a") decisive reason why variation selectors cannot be used with combining marks. In other words, once you try to define <CM, VAR1> as being a variant form of the combining mark in question, you start getting into trouble whenever you try to add another combining mark after it in sequence. > > If CGJ can be used with combining marks in situations where (as far as > we know) there is in fact no problem with multiple combining marks, what > is to stop variation selectors being used in the same situations? Because the situations are different. Apparently you are not grokking this. The umlaut/tr�ma case is one of distinguishing the *collation* order of letters with umlaut versus letters with tr�ma, *not* their appearance. The fact that some minority decides to then also display a tr�ma with a slightly different form than umlaut is besides the point, and does not reflect majority practice even in the German bibliographic data. Furthermore, the recommended sequences here are, I reiterate: <BASE, COMBINING DIAERESIS> (for umlaut) <BASE, CGJ, COMBINING DIAERESIS> (for tr�ma) The CGJ is *not* applied to the diaeresis character -- it is first in the sequence, right after the base letter. The following sequence *is* an allowed one for a variation selector: <BASE, VAR1, combining-mark> as long as the sequence <BASE, VAR1> has *explicitly* been standardized as representing a distinct, graphical variant of <BASE>. If I then apply one or more combining marks to that sequence, there is not a problem. However, such usage defines a variant of the base, not a variant of the combining mark itself. > One > such situation is Holam Male which never takes an additional combining > mark*. So why can't we represent it as <VAV, HOLAM, variation selector>? Because the UTC has ruled out <CM, VAR> as interpretable sequences. > After all in practice there is no normalisation problem with this. (By > the way, I am proposing as one option <VAV, variation selector, HOLAM>, > but that has been opposed on the debatable grounds that what changes is > not the VAV but the HOLAM - the best description is that the whole > grapheme cluster changes.) I don't have a quarrel with describing things that way -- but you just can't get from here to there with variation selectors. --Ken
