CR issues 27-31 (Bjoern's)

Hello QA WG,

We're long overdue dealing with comments that Björn Hörmann sent us in
September, that we have integrated as issues 27 through 31 in our CR
issues list:
http://www.w3.org/QA/WG/cr-issues#x27

They all concern SpecGl ; I believe issues 27 and 28 are moot, due to
the large change of formatting that we envision for SpecGL.
Issue 30 is likely to be solved once the questions regarding the
glossary are closed ; I still think it should remain open as a check
item for the next publication of the new SpecGL.
Issue 31 is editorial ; I think we need to find a better way to ensure
that all the terms we use in Principles and Good Practices are clearly
documented, since they are the equivalent of our old Conformance
requirements ; so this should also stay open as a checkpoint before
publishing SpecGL.

Issue 29 is the most substantial:
"all W3C specifications defining a notion of instance data should be
explicitly required to identify all programmatically reportable errors,
make reporting these a requirement for a specific class of product,
define how to identify such software and define how to identify
instances which do not have reportable errors."

Although it looks unlikely to me that we're going into that level of
specificity in our new guidelines, I think it raises an important
question about error handling in specifications ; the current TR version
of SpecGL doesn't have the word "error" at all in its content, and I
agree with Björn that this topic is important enough that it should be
addressed in the documents ; I had started to noodle on this a while
back in the Wiki:
http://esw.w3.org/topic/ErrorHandling

It's also worth noting that WebArch has also some text related to error
handling, although it's more oriented toward user agents than technology
design:
http://www.w3.org/TR/2003/WD-webarch-20031209/#no-silent-recovery

This also relates to extensibility in some way ; I don't think we're
going to close this issue before the next publication of SpecGL, but I
definitely think it's worth having something in SpecGL about this ; I
would be interested to hear other people opinions on this topic,
especially what are the different error handling mechanisms used across
specifications they know. I would add the summary of such a discussion
in the Wiki.

In any case, I accept to own this issue, and if no real discussion
starts on this thread, I'll try and see if www-qa is more reactive :)

Dom
-- 
Dominique Hazaël-Massieux - http://www.w3.org/People/Dom/
W3C/ERCIM
mailto:dom@w3.org