Sign variation at the handwriting-level was tagged by Simo Parpola in the SAA corpus and this level of analysis was termed by him 'graphetics'. When the Oracc version of the SAA corpus was created, Parpola's graphetics tags were retained in the ATF form '\'[a-z]+, i.e., backslash (reverse solidus) followed by one or more lowercase letters.
Oracc extends this notation to liaise between ATF and OSL in order to allow selection of variant character forms to be integrated into the text stream.
Oracc/OSL does not use the following tags because they are used in SAAo:
\d \dd \ddm \dm \dp \dpm \dt \dtm \dtp \dy \dym \illaja \illaya
\m \md \mm \mp \my \p \pd \pm \pp \ppmm \ppp \t \td \tdm \tdp
\tdym \tm \tp \tt \ttd \ty \tyd \tym \y \yd \ydm \ydp \ym \yp
\yt \yy
Oracc reserves three upper case tags which can be used to specify an Oracc script-style on a per-character basis:
\E - early
\M - middle
\L - late
Oracc reserves mixed-case sequences (upper followed by upper/lower/digits) to allow additional script-style tags to be used, e.g.:
\F - Fara
\G - Gudea
\Gudea - gudea
\ED12 - ed12
Oracc defines some symbolic tags which are mapped to common sets of variations--these are contrived to avoid SAA usage:
\pap - pap-modified form
\k - kaskal-modified form
\l - linear number (all elements in single horizontal row)
\s - sheshig-modified form
\w - wedge (cuneiform) shaped numbers
\v - curviform numbers (not clear that \w and \v are necessary)
These are defined within OSL and the OSL font, and should be implemented in fonts other than the OSL font, with the same semantics, where it makes sense to do so. In the OSL font these are implemented only for a few characters where it is useful to be able to annotate the specific form, e.g., šar₂\v versus šar₂\w; de₂\k versus de₂\s.
In a font which has no characters that vary according to their construction with pap/kaskal/sheshig, there is no need for these modifiers; in a font where curviform numbers are not there is similarly no need to implement \v and \w modifiers.
Oracc defines sequences consisting of 'ss' followed by two digits as accessors for OpenType stylistic set codes, e.g.:
\ss01 - select stylistic set ss01
While early/middle/late are defined by OSL and are in part aliases for the Stylistic Sets defined in the OSL font, the significance of \ss01 etc., is dependent on individual font definitions and thus allows for flexibility to define Stylistic Sets for particular script types outside of the OSL font.
This means that while the sign form accessed by \E in OSL using the OSL font is defined, the signform accessed by \ss01 could vary from font to font.
Oracc reserves numeric tags to access SALT values, i.e., individual character variations:
de₂\1
Selects character variant one for the DE₂ sign in the current font; the same issues of sign form definition apply to SALT tags as to Stylistic Set tags--their behaviour is defined within the realm of OSL and the OSL font, but may vary with fonts outside of that realm.
Graphetics tags may be given on sign values or sign names--the Oracc tools automatically take care of applying them at the proper point in the processing.
The following attributes may be set on an XTF node related to grapheme content--most often on a g:v node:
In ATF the '\' has historically had two distinct meanings. In transliterated texts it is the graphetics tag as described above (formvar). When transliterated words are turned into orthographic forms for lemmatization, formvar tags are dropped by design because the handwriting variants are not considered to be semantically meaningful.
In lemmatization lines, the '\' is used to annotate different grammatical interpretations of a single form (disambiguator):
1. lugal-la #lem: lugal[king]\ak (.ak = genitive) or #lem: lugal[king]\a (.a = locative)
The ATF processor relocates the \ak or \a from the lemmatization onto the form, which results in glossaries/lemmatization tables having distinct forms lugal-la\ak and lugal-la\a and these distinct forms can then have distinct morphology:
@form lugal-la\ak /lugal +-la=l.a #~,ak
@form lugal-la\a /lugal +-la=l.a #~,a
The human interface via #lem lines certainly does not need to change.
While graphetics tags were restricted to SAAo this was never an issue; because @form is parsed using GDL some care may need to be taken to ensure that graphetics/formvars and @form-disambiguators are each understood in their own way. This can probably be controlled at the level of the tool being used; the transliterations tool can assume '\' is formvar; the glossary tool can assume '\' is disambiguator.
If this does cause a problem one possibility would be to change the notation for disambiguators because this is already processed by appending it to the form. The new notation for form-variants could be '\\', for example, or \=: @form lugal-la\=ak, say. It may not be a problem, though.