pkgsrc-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: ANN: Availability of pkg(8)-capable pkgsrc

I'm not looking to enter the political side of this, but I will say this as a user. I use FreeBSD and NetBSD daily. I use ports and pkgsrc, almost daily. Today, for this particular instant, I find 'pkg' slightly better, though I really appreciate pkgin as well. Here are my thoughts:

* The name is less to type and makes more sense. It's stupid, but renaming
  it 'pkg' or 'pk' would be positive, IMHO. I actually wrote a wrapper
  script to front end pkgin, pkg, apt-get, yum, and others with "pkg" (my
  front end script name).

* The 'pkg' tool has more functionality than 'pkgin'. For example, pkgin
  can show you a list of all installed packages with "pkgin list" but
  FreeBSD's can do that _and_ take an argument in the mix (-l) to show the
  files in the package. NetBSD's can't do that. You have to go use
  traditional pkg_info (-v) or another method for that.

* I sometimes notice that the network/index/repo files pkgin wants are
  either not there or are massively out of date with the packages. I
  notice that pkgin doesn't seem to check, either, wheras pkg does this
  every time I do any kind of repo-based action. Perhaps I'm wrong about
  this, and maybe it's just that pkgin doesn't pull as many updates
  because there aren't as many to pull. However, I seem to need to
  manually 'pkgin update' a lot when I notice it's failing to grab

* The FreeBSD 'pkg' tool seems to use fetch(3) which allows for more
  user control over it's network dynamics. whereas the NetBSD one uses
  some less flexible internal fetching logic that just locks up and falls
  apart when there are network problems. It locks up and has to be killed
  sometimes when it's in that state. I haven't seen this happen in a few
  months, though. Maybe it was fixed or maybe it just coincided with my
  getting an upgraded internet connection (which is what I suspect). Also,
  I never tried a kdump to see if it did any getenv() calls for another
  fetcher. I could be all wet here if there is an option I don't know of.

* Neither of them can give you a way to emulate yum's "whatprovides"
  feature. That's a huge miss for a sysadmin type guy like myself. Yum/dnf
  is a total turd (rinky dink script, actually), IMO, but that feature is
  an absolute godsend since many times package names don't include the
  command names, and that's what folks memorize.

Overally I'd simply say that FreeBSD's 'pkg' (pkg-ng) is baked, but pkgin is still in the oven. There several smart folks who participated in creating NetBSD's 'pkgin'. The items on this list are pretty trivial, really. They could easily be addressed and the last one I mention may already be fixed for all I know. I feel like the pkgsrc project has a more utilitarian world view than any other projects of it's kind (Free/Open BSD ports, DragonFly ports, OpenPKG, etc). I want it to have the best in everything and I'm not bashing it. I'm just making an eyes-wide-open set of observations on pkgin (which is a nice tool that I'm grateful for) as a passionate user of all things BSD.

I'm not in a snit about either one. That's for sure. I'm glad for both.


Home | Main Index | Thread Index | Old Index