[gentoo-dev] Re: [soc] Python bindings for Paludis

[Duncan]

Mike Auty <ikelos@g.o> posted 460FAFF4.4060901@g.o,
excerpted below, on  Sun, 01 Apr 2007 14:13:24 +0100:

> From what I've read of the PMS, it currently only describes the input
> format it would accept (namely the format for ebuild files and their
> contents).  This question can be delayed until the PMS defines the
> operation of the package manager, including but not limited to the
> recording of installed package data.  If the package managers do not
> agree on which packages are installed or how to uninstall them, then
> they are not yet interchangeable.
>
> I apologize if this point has already been raised elsewhere in the
> thread.  I try not to get involved in threads like this, but
> accidentally read a reply and thought this might be a valuable response.

Thanks.  It is valuable (and hasn't been already raised to my 
observation).

As I understand it, they wouldn't necessarily be dynamically 
interchangeable, that is, on a live system (at least not without running 
some sort of conversion utility, which may or may not exist and may or 
may not "lose" some information in the conversion, defaulting the missing 
values).  Rather, one could choose one and run with it, and only change 
with some work and/or loss of data.

Practically speaking, at minimum, it is assumed the world file would 
normally either remain the same format or be convertible (automatically 
or by hand), and the USE flags would be convertible, so if install data 
were lost in the switch, one could at worst rebuild empty-tree world (in 
whatever PM lingo) to get the database in the new format if necessary.  
Thus, it's not something one would wish to switch back and forth willy 
nilly, but switching would be possible, with a bit of work.

Of course, that assumes a package manager that even has the concept of 
the world file, I'd guess a /relatively/ safe assumption (and some form 
of USE flag handling is required by the spec).  For those that didn't, 
well, a rather more painstaking individual package rebuild may be 
necessary to do the conversion.  However, one might suppose those would 
be corner cases, and if someone wanted to go to the trouble, well...

The point being, however, that all this quarreling about "official" 
package managers doesn't /really/ have to happen.  Arguing Ciaran's 
viewpoint for a moment, if portage really is /that/ bad and "future 
challenged", if official restrictions /do/ end up eliminating all other 
competition for official manager, well, it's entirely possible there'll 
be an official manager, and then there'll be the one (or more) everyone 
actually uses, again making arguing over an "official" PM "much ado about 
nothing".  That's one extreme.  At the other, the alternatives just never 
go mainstream, regardless of whether they are "officially blessed".  
Again, much ado about nothing.  In the middle, multiple managers prove 
functional and are chosen by a reasonable segment of Gentoo users, 
regardless of "official blessing" or not, and again, it matters little 
what the official status is.  I just don't see why so many are spending 
so much time arguing over it, when regardless, people are going to make 
their choices, and "official" status won't matter so much when people do 
so, because the package spec and what works is going to be the defining 
factor, not some "official" blessing, except indirectly as that affects 
further package spec updates.

If that makes any sense and isn't entirely circular... it does (and 
isn't) to me, anyway.  Certainly more so than what to me is pretty much 
bickering over nothing.  =8^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

-- 
gentoo-dev@g.o mailing list