First of all, see this blog entry for a not-quite-comprehensive review of the mass chaos in multiple-maintainer changelogs that existed when dch was modified to try to support them:
http://kitenet.net/~joey/blog/entry/devscripts-2005-07-02-22-56.html Note that the format ("parens format") historically used for the Debian kernel changelogs is not machine parseable. There is no way to tell if two words in parens at the end of a changelog entry are an author name. Therefore dch cannot support it. Also note that the dch manpage documents how to turn the multiple maintainer mode off, so I hope you're not serious in <[EMAIL PROTECTED]> about forking dch. > First, it emphasizes the author, and not the actual change. This is a > changelog, after all; the person who actually did the change should come > after, not before. That leads to a constuct that looks in general something like this: <description of change> <author name> <description of change> <author name> -- <releaser name> Whether the names are the same or different, this looks weird because you have two names mashed up next to each other. I don't entirely understand your objection to putting the author's name first, since it's easily ignorable or easily read if you like to read who's saying something before reading what they say. Which does increase overall reading comprehension with some authors. ;-) Heck, it works on irc.. > Second, the indentation makes it hard to scan through; the author and > description lines are aligned. > For listing changes grouped by author, it would still be nice to see > something like the following (keeping consistency w/ older changelog > entries that start w/ *): > > * [ Julian Gilbey ] > - bts: fix clone command (Closes: #321798) > - debchange: add comment to manpage about building sponsored uploads > (Closes: #321960) > - debchange: introduce --create option: allows creation of > changelogs / NEWS files (Closes: #220755, #322716), and > inform user about this option if changelog / NEWS file is > not found (Closes: #316661) This seems to emphasise the author (and thus, impede scanning through) much more than the braced maintainer format, and it also wastes a lot of horizontal space and leads to uglyness when changes need nested indentation. We used this in d-i for a year (except the braces around the author name), and it rather sucked. My blog entry lists 5 serious problems with it. Also, even with the braces, it is not machine parseable; there is no particular guarantee that two words in braces at the start of a changelog entry is an author's name. > Third, you end up w/ a lot of repeated author entries. > However, the important thing is for chronologically entered changelog > entries to not become cluttered with author names... Speaking of clutter, the old kernel changelog format was asoundingly cluttered with author names, as they appeared at the end of each line and were in parens the same way bug closures were which impeded mentally filtering them out. My experience as a reader.. I suspect that if you go back and reformat it using brached maintainer you will see a net reduction of author names, a reduction of number of lines used by the changelog, and a reduction in bytes used. > It would be really nice if there was a way to distinguish whether you > wanted dch to list entries chronologically, or grouped by author. > In the case of being grouped by author, the current format could be used. > In the case where you want to list entries chronologically, perhaps the > following format could be used > > * Change ATM and Classical-IP-over-ATM to be modular, instead of being > statically included. > (closes: #323143) > [ Bastian Blank ] Please bear in mind that this has to be machine paresable for dch to to its work. Switching back and forth between two formats, especially since both will invariably show up in the same changelog entry, will not aid machine (or human) readability. Also, the format you've proposed here is not particularly machine parseable, since there's no reason a changelog entry cannot end with something in braces that is not actually an author's name. I am open to machine parsable formats that are better than the current format though. I think dch should support any reasonable competing formats until there's a clear consensus. And Debian does need to come to a consensus, the plethora of formats used for one or two packages each was making apt-listchanges ugly and hard to read. So far the only other contender I know of, what I call "dash format" in the blog entry, does not seem likely to please you either. -- see shy jo
signature.asc
Description: Digital signature
