L2/03-029

Subject: Suggested changes in UTC / L2 procedures
	    Concerning: Reviewing WG2 drafts and Unicode properties

Source: Asmus Freytag
Date: 01-30-2003

Background:

In publishing AMD 1 and 2, we found four separate errors in both names and  
glyphs. (For glyphs these were identity issues, i.e. the right glyph in  
the wrong place). The name errors were all mis-spellings of one sort or  
other. The last of these (so far) was reported today. It would be nice if  
we could say that these errors only relate to last minute additions to the  
amendments in Tokyo. This is NOT true: most of them go back to the first  
PDAM. This is an indication that our process for reviewing WG2 drafts is  
seriously broken.

The seriousness of this situation cannot be stressed enough. In  
consequence of our stability policies, some of these errors cannot be fixed  
after publication.

Proposal

In order to improve our process I suggest that either UTC or L2 formally  
appoint one or more individuals whose agreed upon task it is to review a  
specific script or range of proposed characters in each draft for accuracy.

Expertise in a given script is not the requirement, as most of these  
errors can be found by systematically correlating the original proposal  
documents and resolutions from the WG2 minutes with the corresponding  
sections in the PDAM, or FPDAM.

The editors are both too close to the work and too wrapped up in  
production details (such as managing the mechanics of files, fonts and  
documents) to have the time or the distance from the work to be able to do  
this review. It would also add a significant amount of work.

However, if this task can be spread across a number of people, each with a  
well-defined action item to report to UTC on a specific subset of  
characters, the workload would be quite manageable and the people doing the  
checking would not be subject to being lured into false security from  
having seen the mistaken version over and over again.

The same procedure could be applied to reviewing the Unicode character  
properties, by breaking the task into chunks and giving specific people an  
action to review their chunk (and anything else they feel like, but at  
least their chunk).

Relying on this to happen spontaneously, or relying on Ken to do it all  
does not produce quality results.