On Thursday, January 12, 2006 02:07:29 PM -0800 Dave Crocker
Suppose that we edit the document in XML (we are already
doing this part of the time), do a final nroffing pass to get the
format just right, and then give the author(s) the edited xml,
final .txt, and a diff file. (We could easily do this today).
The author(s) change the .xml (or give us a patch file for us
to make the changes.) We have to make a new nroff version and
PUT THE SAME CHANGES INTO IT THAT WE DID THE FIRST TIME.
This sound like quite a good approach.
It sounds like an awful waste of time and effort to me.
It seems like the more efficient approach would be to essentially have two
stages, where the authors first sign off on the result of copy-editing, and
then on whatever cosmetic changes are needed after the final conversion.
In the few cases where I've had the chance to observe the AUTH48 process on
a large, complex document (mostly as a WG chair), the majority of issues
have been related to changes in wording the authors were unhappy with,
missed spelling errors, etc. If these issues were resolved using only the
XML and the output of xml2rfc, with the nroff and final text not being
generated until the authors were happy with the copy, I'd bet that 95% of
documents would result in no additional comments from the author based on
the final text.
Ietf mailing list