Re: QAWG position on errata proposal

I found Ian's proposal of last year:
http://lists.w3.org/Archives/Member/chairs/2001OctDec/0037.html
which I like, especially the part that says: "The only 'normative'
corrections to a Recommendation are those in the Approved state."
This implies that errata on their way to being approved are
visible.

There is a newer proposal from Ian and C.M. Sperberg-McQueen at
http://www.w3.org/2002/05/31-errata
Feedback requested at
http://lists.w3.org/Archives/Member/chairs/2002JulSep/0125.html
It has a good definition of "substantive changes" tied directly to
the conformance implications. It proposes a heavyweight process for
such changes, which is good in the sense of providing a public
review period, but risky if it discourages WGs from dealing with
errata in a timely fashion.

The proposal includes an erratum mailing list for each Rec, to which
an interested person could subscribe. There is a large gap in the
proposal, though: it is only used for notices of proposed and maybe
approved errata, but it should also be used to announce that an
issue is on the "errata track" because the process is so heavy. For
example, when I send in my issue about empty lists and xsl:number
to the XSL WG, I think that xslt-errata subscribers should know
right away, so that they can send in test cases that expand the
analysis of the issue. If others don't learn of it until the WG has
boiled an erratum out of my issue, their response will come at a
late stage, and there is a much higher likelihood of the erratum
having to go back for revision. (BTW, the issues I send arose in
response to an approved erratum that was not as complete and final
a disposition as the WG evidently thought it was.)

The proposal suggests formal publication no more often than every
three months, so earlier public notice allows implementers to bring
their products in line with proposed errata that seem likely to be
approved. Test case writers can get tests ready, particularly the
errata annotations, so that they can be added to the suite on the
day approval is announced.

Errata would "become normative only when republished on the TR page."
This is a possible problem if the time lag gets too long. Other than
that, I'm favorable. Of course, we also need to have timely response
for Recs that are actively used but whose WGs are disbanded.

Errata should be numbered, so that test case exclusion can be driven
by a consistent ID scheme. It helps for citations, too.
.................David Marston

Received on Tuesday, 8 October 2002 15:39:00 UTC