Re: [UDD] Package file lists / normalization
On 20/04/13 at 16:43 +0200, Michael Stapelberg wrote:
> Hi,
>
> I was looking into importing package file lists into UDD today and got a
> proof of concept working.
>
> The following questions arose:
>
> 1) This data takes up quite a lot of space, i.e. 276 MiB for amd64 only.
> Is such an increase in space acceptable for UDD? If not, why not?
>
> 2) In order to save space, it’d be reasonable to normalize the data,
> i.e. not store the same value over and over again. This is currently
> not done anywhere in UDD. Was this a conscious decision? If so, why?
>
> I suggest storing the package file lists in a normalized way and
> providing views to access them in an unnormalized way, if the user so
> desires. What do you think?
>
> In case you are wondering: my use-case to have these lists in UDD is to
> create an analysis of the different init systems in Debian, e.g. how
> many packages ship a sysvinit script and what percentage of those also
> ships a systemd or upstart file. In case you know an easier way to get
> unpacked package file lists, please let me know.
Hi Michael,
There has been some discussion on the lintian side to add some
"feature" tags, that would not be things to fix, but rather information
about the package.
I have the impression that doing such an analysis on the lintian side,
and then getting it imported into UDD through the lintian importer "for
free", would be a better design.
What do you think?
Lucas
Reply to: