Atlas proposal [and 1 more messages]

Ian Jackson ijackson at chiark.greenend.org.uk
Thu Aug 26 10:52:16 UTC 2010


sean finney writes ("Re: Atlas proposal [and 1 more messages]"):
> as an admin i'd be very annoyed by that behavior, because there's no way
> for me to know what package the file came from then, or whether it was my
> own accidental actions that led to it (i.e. a make install gone wrong
> somewhere, etc), and no way to do checksum verifications.  

There must be many other things that make you very annoyed, then,
because there are loads of files in /usr which are created by package
maintainer scripts.

If this kind of thing bothers you then you should be pursuing those
file-registration schemes so that you can look up the origins (and if
you like, checksums) of files more easily.  If there were a standard
file registration scheme for files which were in the .deb's filesystem
archive, naturally the proposed atlas installer would use it.

> to me it seems that somewhere under /var/lib or /var/cache would be
> more appropriate...  and just to throw something out there, there is an
> /etc/ld.so.conf.d directory, so the package could drop a file in there
> pointing at wherever the libraries were placed.

Yuck.  This is complicated and unreliable.  The difference between
/var and /usr is that /usr is not modified _except at package
installation/upgradwe time_, not that /usr isn't modified at all,
obviously.

Ian.



More information about the debian-science-maintainers mailing list