Portage Directory Structure

[rmsousa]

I can't understand portage's directory structure.
by "cache" I understand something that can be deleted without harm, as it only saves you some downloads
And then I come across these things:

/var/cache/edb, which contains, among others, world, which CANNOT be deleted, as it is a configuration file (should be in /etc?)
/usr/portage/distfiles, which contains files which I can download at any time (should be cache?)
/usr/portage/x-y, which contains the ebuilds, that, again, can be downloaded again with an emerge sync (okay, a LOOOONG emerge sync, but anyway...)

That is : everything that is in /var/cache/edb/ should be in /var/lib/edb or /etc/edb
everything that is in /usr/portage (except profiles, eclass, scripts, etc) should be in /var/cache/portage

So, if I run _really_ low on HD space, I can just delete /var/cache, which, in my opinion, is expected behaviour.

Can someone enlighten me on why it isn't organised in this way (or a similar one)?

[David_Escott]

I think the cache here is refering to the fact that this enables portage to make lookups faster so it is in that sense a cache (just as a cpu cache is faster access than going to ram). I'm not entirely sure where portage does keep all its databases but the world file is in fact not an important file.

One thing to note is that distfiles should truly not be deleted in total like that. rather you should just delete the ones you dont need anymore. There are some scripts in the forums that do this.

[axxackall]

As far as I remember Unix, /usr was used to be a filesystem often mounted as read-only to keep all installed software files if they are not changeable between software reinstallations. As read-only it's less likey will be corrupted at system crash. It was remounted as writable only when a sysadm reinstalled something. Thus, /usr/portage was a good choice - it's a place where portage tree is installed. You reinstall the portage tree by "emerge rsync". However, re-installation of portage tree is not an exception, but rather a normal daily (at least weekly) based operation in Gentoo. Thus, if you resync often then it's a good idea to move /usr/portage to /var/lib/portage

/var in Unix was for changeable data, like databases. Like /var/cache/edb, which mostly keeps cached deps. /var/db/pkg keeps the database of currently installed packages. They are changed not because you resync the portage tree, but b/c you want to (re)install some package from the portage tree, after resync or after you just decided you need a new (version of) package. Means you do it much more often than you resync /usr/portage.

Let's thing about /var from DBA's point of view. Would you delete your Oracle or PostgreSQL DB when you have to free some space? No, you would not. Instead, you run sqlplus or psql and delete some records. Or you clean DB with some "defragmentation/vacuum" scripts. Same thing, you run some scripts which will clean up /var/cache/edb.

I agree /var/cache/edb is not exactly DB as you can rebuild it again anytime. Just it will take lots of time - that's why don't move it to /tmp or your boot will take really long time.

/tmp was for files which you can delete anytime. What's even more important, it's supposed to be cleaned up automatically after every reboot. However, I would not put any ebuild temp folders into /tmp as I may need them again after reboot. Otherwise, it might not a bad idea to move /var/tmp/portage to /tmp/portage