Subject: Re: pkgsrc progress bar?
From: Dmitri Nikulin <email@example.com>
Date: 01/22/2005 02:44:31
Sorry, I must have misunderstood. It's when you inquired about a 'time'
estimator that I figured you were after an absolute scale.
But in the sense of package/packages, yeah, that's another area where
Portage shines. pkgsrc is still, if you look at it, an admittedly
elegant hack of package management, but writing a new package manager
based on a tool chain (like Portage is) and with more advanced design
would be really useful. I hope the DragonFly system develops this way.
USE flags, pretend mode with an exact list of packages to be merged and
how much will be downloaded, tested/untested separation, transparent
integration of distcc, ccache, etc. are all really handy, and the only
reason I run Gentoo on workstations instead of the technically superior
NetBSD. Package management makes or breaks a distribution these days
where most people expect convenient user lands.
package/packages is possible to implement as a hack to pkgsrc, but
whether or not it's worth complicating things for this is questionable.
The simple "I don't care about what happened before or will happen
later" tree design of pkgsrc/ports/etc is admirable from a development
perspective but it becomes a pain when you want to predict behavior and
My view on it probably will offend some more experienced developers,
which is fair enough: but as a user of FreeBSD, NetBSD and Linux (with
Gentoo being the only distribution I'll even look at) I know what a good
package manager is from an end-user and administrator perspective. It's
one area that could bring NetBSD from being an OS for power users to
being an OS more for everybody; and while most won't ever contribute
back, eventually the spread of "hey, it's convenient too" rumors will
attract developers looking for a marketable platform. Linux is a toy
that shouldn't be used outside of a play pen, but entirely because of
good marketting and efforts towards meeting the needs of every Joe User
and Joe Loser, it's become a majorly supported and distributed toy.
There is no reason that NetBSD shouldn't have as much if not more
attention. And, yes, a simple a thing as a package manager would be a
huge leap. I could develop the tools myself; that's not the hard part.
The hard part is getting a large scope of enthusiasts to contribute and
maintain the 'ports' of software. And the easier you make that task, the
more the system succeeds. So a better system helps users and developers,
and the host operating system (but an ideal package manager would be
portable like pkgsrc is already).
(and sorry for cross-posting this, but I want to ensure those reading
the old thread in current-users@ know where to track it)