[Latingp] Digraphs

Dillon, Chris c.dillon at ucl.ac.uk
Mon May 16 08:39:58 UTC 2016

Dear Meikal & Abdeslam,

Thank you for your emails. This correspondence is a good summary of answers to difficult questions, along these lines:

  *   Variants may consist of more than one code point.
  *   So far we have been able to exclude combining marks, but it is doubtful that that will continue to be possible once more work has been done on the use of the Latin Script in Africa. I would suggest that we take the approach "combining mark X is required in the following sequence(s) of code points only", rather than "combining mark X is included with any other code point".
  *   As regards ij and most other ligatures, they would be unallocatable variants, or possibly out-of-repertoire code points.
  *   I like the suggestion of waiting for the IP's informal comments before releasing our draft repertoire. The Second Level Team's work, however, could require a substantial effort to digest and so we should probably wait.
Français: Ces emails forment une synthèse utile de réponses à quelques questions compliquées:

·         Les variants peuvent consister en plus d’une lettre Unicode.

·         Si on a besoin de signes pour combiner des lettres Unicode, on pourrait seulement les utiliser en des cas limités.

·         Ij, etc. sont peut-être un variant de i + j qui ne pourraient jamais exister dans un TLD, ou bien peut-être tout à fait hors de notre répertoire.

·         On va attendre seulement jusqu’à ce qu’on ne reçoive les comments informels du IP avant d’inviter des comments sur notre répertoire.



On 14/05/2016 10:50, Meikal Mumin wrote:
Dear colleagues,

so that clarifies that question - thanks Abdeslam.

Coming back to your questions Chris - I believe combining marks could be excluded, as was done in the case of Arabic LGR. Meanwhile case like ij could be declared variants with a sequence of i + j, provided we see a need for including the former.

If ligatures are no part of MSR-2, then I assume the problem has solved itself.


Dear colleagues,

I would suggest waiting for the feedback from IP, but not for anything regarding second levels.



2016-05-11 22:27 GMT+02:00 Abdeslam Nasri <abdeslam.nasri at gmail.com<mailto:abdeslam.nasri at gmail.com>>:
Dear Chris and Colleagues,

Digraphs or more generally sequences of code points, can be specified as variants of a single code point.

An excerpt from the LAGER specification :

" A sequence of multiple code points can be specified as a variant of a

   single code point.  For example, the sequence of LATIN SMALL LETTER O

   (U+006F) then LATIN SMALL LETTER E (U+0065) might hypothetically be

   specified as a variant for an LATIN SMALL LETTER O WITH DIAERESIS

   (U+00F6) as follows:

       <char cp="00F6">

           <var cp="006F 0065"/>


In the typical case of digraphs these are named precomposed versus decomposed formats of a single letter. Normalization should exist in Unicode in order to allow these variants, or otherwise block them.

Kind Regards,
Abdeslam NASRI

2016-05-09 15:43 GMT+02:00 Dillon, Chris <c.dillon at ucl.ac.uk<mailto:c.dillon at ucl.ac.uk>>:
Dear Meikal,

Thank you for your thoughts on digraphs.

In that case, we would have blocked variants like i, dotless i  and iota, where application for a label containing one, would block applications for labels containing any of the others.

We would also have blocked variants, digraphs like ij, which could never be allocated at all. If we need to do this, it will be necessary to describe variants for ligature code points we have not yet analysed in the Latin ranges, as they aren’t in MSR2.

(This distinction is what I was finding difficult during the face-to-face meeting in Marrakech.)

Incidentally, I’m fairly sure two code points could be a variant of one. ( I wonder what happens with the Arabic ligature of laam and alif that looks like Greek gamma; in Urdu the two do not combine so closely, if at all.)


Research Associate in Linguistic Computing, Centre for Digital Humanities, UCL, Gower St, London WC1E 6BT Tel +44 20 7679 1599<tel:%2B44%2020%207679%201599> (int 31599) www.ucl.ac.uk/dis/people/chrisdillon<http://www.ucl.ac.uk/dis/people/chrisdillon>

From: Meikal Mumin [mailto:meikal.mumin at uni-koeln.de<mailto:meikal.mumin at uni-koeln.de>]
Sent: 09 May 2016 09:38
To: Dillon, Chris <c.dillon at ucl.ac.uk<mailto:c.dillon at ucl.ac.uk>>
Cc: latingp at icann.org<mailto:latingp at icann.org>
Subject: Re: [Latingp] Digraphs

Dear Chris and colleagues,

apologies for the late reply. I believe we don't need to exclude digraphs. We could simply set them up as variants, e.g.  ij as equivalent of i + j. It could be useful to verify with IP, if it is possible to declare a sequence of two code-points as a variant of one - we had not encountered such a case with Arabic script.

Best wishes,


2016-03-29 9:54 GMT+02:00 Dillon, Chris <c.dillon at ucl.ac.uk<mailto:c.dillon at ucl.ac.uk>>:
Dear colleagues,

Mirjana’s recent research on Montenegrin has raised some interesting issues.

One of them is diagraphs.
Currently we have digraphs like æ and œ in our repertoire, but Dutch ij (U+0133) as in vijf ‘five’ is white in MSR-2 (not compatible with IDNA 2008). Certainly many digraphs, including ij are visually similar to their component letters. We could consider adding all digraphs to the list of criteria for exclusion, or adding them with exceptions (less good from a usability point of view). Incidentally, ß and & are probably excluded for other reasons, Longevity Principle and Punctuation, respectively.

What do you think?

Français: Qu’est-ce qu’on devrait faire avec les digraphs dans notre répertoire – les permettre ou pas?



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mm.icann.org/pipermail/latingp/attachments/20160516/75ab2c68/attachment-0001.html>

More information about the Latingp mailing list