[pacman-dev] package PKGINFO format vs DB format

[Roman Kyrylych]

2007/6/26, Dan McGee <dpmcgee at gmail.com>:
> On 5/17/07, Roman Kyrylych <roman.kyrylych at gmail.com> wrote:
> > 2007/4/23, Aaron Griffin <aaronmgriffin at gmail.com>:
> > > I may be missing some history here, but I am wondering why the package
> > > metadata is in a different format than the DB metadata.
> > >
> > > As far as parsing goes, it would be ideal to unify the two formats.
> > >
> > > Does anyone (judd, you're CC'd on this) know any reason to keep two
> > > different formats, or any tools that would be negatively impacted by
> > > such a change? Please note, a "negative impact" is not that same as
> > > "we have to change it for the new format" - more along the lines of
> > > "we can no longer do XYZ".
> >
> > Status?
> > I don't know any reason to keep them different.
> >
> This died a LONG way back. But I have something relevant here,
> courtesy of Chantry:
> http://www.archlinux.org/pipermail/pacman-dev/2006-March/005702.html
>
> I'm all in favor of doing this, although it is going to break some
> things. We will need to plan a transition roadmap, especially if we
> want to get the local DB in the same format.

+10 for that new format!
(IIRC it was mentioned at least once after that date too)

I see two possibilities:
1) leave the current code for handling directory-based database during
the transition period,
2) modify it to do a conversion to a new format when unpacking
.db.tar.gz or finding existing unpacked database in old format.

.db.tar.gz could be kept in current format for a while, to allow
people who upgrade rarely to avoid problem.
Or we can start using new format quickly, but leave
pacman/{desc,depends} in a db, which will be parsed by old pacman, but
not a new one. This will allow users to upgrade pacman first, then all
other updates can be fetched by a new pacman.

-- 
Roman Kyrylych (Роман Кирилич)