Re: Making it easier to get and use pkgsrc

[Aleksej Saushev <asau%inbox.ru@localhost>]

David Holland <dholland-pkgtech%netbsd.org@localhost> writes:

> On Sat, Dec 11, 2010 at 08:14:26PM +0000, David Brownlee wrote:
>  > It has been suggested that we should make it easier for users to
>  > download and start using pkgsrc.
>  > [snip]
>
> All this discussion seems futile and rather silly, because it seems
> fairly obvious that the correct approach is to have sysinst offer to
> download and unpack /usr/pkgsrc, either stable or HEAD as desired.
>
> This should be straightforward, really.
>
>  > - Default to DEPENDS_TARGET=package-install and
>  > UPDATE_TARGET=package-install
>
> as pointed out already, this is wrong. (and we can't switch to
> USE_DESTDIR by default, btw, until someone fixes undo-replace. See PR
> 43689.)
>
>  > - Move PKG_DBDIR under PREFIX
>
> This is a bad idea. As I've said before, PKG_DBDIR contains exactly
> the sort of data that is supposed to live in /var: mutable indexes and
> databases that reflect the state of the system.
>
> Surely you don't also think it would be a good idea to move the
> comparable base-system data from /var to /usr? (e.g., the locate
> database, services.db, /var/db/obsolete, and so forth.)

I disagree. There's 1-1 relationship between PKG_DBDIR and LOCALBASE,
if you have several pkgsrc deployments (some do utilise this in complex
administrative schemes), then you have to watch not only for
subdirectories in /usr, but also for subdirectories in /var.

Using /var for _really_ variable data would be more helpful.
Good example for these data is PostgreSQL database, which is created
by strange decision under PREFIX by default: PGHOME?=${PREFIX}/${PGUSER}
(see postgresql*-server/Makefile). Note, that there's usually only one
PostgreSQL server running, if you want to run more servers, you have to
tune them anyway (PostgreSQL creates socket in /tmp).


-- 
HE CE3OH...