Apparently, though unproven, at 22:45 on Saturday 05 February 2011, Volker Armin Hemmann did opine thusly:
> again, you are starting from a mistaken premise. > > /usr/portage makes sense, when you consider its history. It may not be the > appropriate decision, but with its background it was logical back then. > > And if something is not broken, don't change it. You do not know what old > tool/setting/whatever might suffocate. > > PORTDIR is not a mere workaround. If you are sure that there is no old crap > lingering around that might expect portdir as /usr/portage, use it. > > Besides /usr/src/ contains linux and other sources. Wrong too? It is f* > tradition. portage does not contain temporary data or database stuff - that > crap is in /var/db, /var/tmp/portage, /var/lib. So the worst stuff is > somewhere already. Tradition on it's own is a lousy idea for retaining anything. A tradition worth keeping is one that's worth having because it has use. However most traditions are merely "but we always did it this way..." /usr/portage is a tradition, a hangover from BSD. LFS is a standard and /usr/portage gets in the way of the standard. Guess which one should trump the other? And the portage tree IS a database. You put (or cause to be put) data into it, which can be amended, edited, added to or removed, other actors query the database for information (emerge, eix, etc). The fact that it is updated on demand and not on the fly, that it is not relational in nature, that it doesn't have "sql" anywhere in it's name and that it is purely file-based does not detract in the slightest from the simple fact that the tree is a database. It's just plain outright stupid to have a default location for something (that by definition is variable) in a place that by definition (or by de-facto consent) must be mountable read-only and have no ill effects on the rest of the machine. -- alan dot mckinnon at gmail dot com
