[gentoo-dev] gentoo's portage

[gentoo-dev] gentoo's portage

[Sami Cokar]

Hello,
   I've started playing with gentoo 1.0r6 and have a few comments.  I
installed gentoo using the tbz2 file and have a few comments:

1) documentation:  please ensure the 'build' options
(/etc/make.conf.build) are explained as to what all the three letter
acronyms mean and the impact/dependencies of picking them

2)documentation:  the departure from 'normal' system startup scripts
needs to described.  What will happen when a 'normal' non-gentoo
application tries to install itself expecting a 'standard' layout?

3) portage:  a great drawback seems that 'constant human effort' is
required to ensure the portage tree is updated.  I can see this becoming
an issue for gentoo.  There are many applications released and updated
that must conform to the portage system to make sure 'the latest and
greatest' really IS available "as advertised" on the web site :) eg.  I
needed the latest Matrox drivers and Powerdesk utility which I installed
manually.  my system's portage is not aware of manual installs I perform
and I needed the drivers before they could be added to the 'gentoo
portage' tree.  

4) portage: sounds powerful - perhaps it can also be used as a means of
a useful/automated/'user-friendly' 'uninstall' which is sorely lacking
in the unix world.  I have tried 4+ distributions of Linux and am
getting a bit tired of telling their respective 'uninstall' that when I
want to remove a game, I would STILL like to keep gnome on my system.

5) portage: if possible (space requirements, etc), 'generic 386'
binaries for the base system should be in the portage tree so that
initial installs/builds can progress faster.  the 'emerge system' on my
P4 1.4 took a looong time.  Would be better to have as much generic base
system and then be able to fine tune as desired.

6) portage: organization.  consider reducing the # of categories.
People can mainly remember 7 +/- 2 items of information.  go into
/usr/portage and there are a lot of choices which will only increase in
the future.  Consider:

    /usr/portage/base  (anything necessary to get from PC booted to
minimal command line)
    /usr/portage/GUI   (user interfaces, windowing systems)
    /user/portage/development  (development tools, languages)

    etc.    Note that i used 'development' instead of 'dev'.  lets use
long file names as necessary.  

    and don't go too deep:
    
 
/usr/portage/development/languages/python/libs/3rdparty/UnimportantLib



So far, gentoo seems to be the distribution I would prefer to use.
Congratulations on the results you've achieved.

Thank you.

____________________________________________________________
Sami Cokar            Landmark Graphics       scokar@lgc.com
Phone: 403-218-2297                        Fax: 403-233-0141

Re: [gentoo-dev] gentoo's portage

[Damon M. Conway]

>1) documentation:  please ensure the 'build' options
>(/etc/make.conf.build) are explained as to what all the three letter
>acronyms mean and the impact/dependencies of picking them
>
>2)documentation:  the departure from 'normal' system startup scripts
>needs to described.  What will happen when a 'normal' non-gentoo
>application tries to install itself expecting a 'standard' layout?

Yes, we need documentation.  It's probably about time we look at putting a
doco team together.  The 1.0 release is sneaking up on us.

>3) portage:  a great drawback seems that 'constant human effort' is
>required to ensure the portage tree is updated.  I can see this becoming
>an issue for gentoo.  There are many applications released and updated
>that must conform to the portage system to make sure 'the latest and
>greatest' really IS available "as advertised" on the web site :) eg.  I
>needed the latest Matrox drivers and Powerdesk utility which I installed
>manually.  my system's portage is not aware of manual installs I perform
>and I needed the drivers before they could be added to the 'gentoo
>portage' tree.  

Please see the 4000 packages in the FreeBSD ports tree.  This method is
tried and true.

>4) portage: sounds powerful - perhaps it can also be used as a means of
>a useful/automated/'user-friendly' 'uninstall' which is sorely lacking
>in the unix world.  I have tried 4+ distributions of Linux and am
>getting a bit tired of telling their respective 'uninstall' that when I
>want to remove a game, I would STILL like to keep gnome on my system.

It does, but it's not very userfriendly at the moment.  To unmerge an app,
do this:

ebuild /var/db/pkg/<category>/<app-ver>/<app-ver>.ebuild unmerge

It will only unmerge that version of that application, and nothing else.

>5) portage: if possible (space requirements, etc), 'generic 386'
>binaries for the base system should be in the portage tree so that
>initial installs/builds can progress faster.  the 'emerge system' on my
>P4 1.4 took a looong time.  Would be better to have as much generic base
>system and then be able to fine tune as desired.

Yes, but right now we are in a pre-release state and doing a full build for
each release would get us nowhere.

>6) portage: organization.  consider reducing the # of categories.
>People can mainly remember 7 +/- 2 items of information.  go into
>/usr/portage and there are a lot of choices which will only increase in
>the future.  Consider:

They don't have to remember them, they are in /usr/portage.  They are also
pretty self explanitory.

>    /usr/portage/base  (anything necessary to get from PC booted to
>minimal command line)
>    /usr/portage/GUI   (user interfaces, windowing systems)
>    /user/portage/development  (development tools, languages)
>
>    etc.    Note that i used 'development' instead of 'dev'.  lets use
>long file names as necessary.  
>
>    and don't go too deep:
>    
> 
>/usr/portage/development/languages/python/libs/3rdparty/UnimportantLib

This is no better than  dev-python/libN.  You have to go no farther than
/usr/portage to get a view of where things are.  Also, you can do an
"emerge <category>/<app>" from anywhere in the filesystem, and portage with
install the latest version of that app with dependencies.  Can you imagine
having to type "emerge
development/languages/python/libs/3rdparty/UnimportantLib" ?

I think the more you play with the system, the more you'll begin to see 
it's logic.  Also, I know that some of us would like to see a search 
feature.  However, that is low on the list for now.  We're in the process 
of shaking bugs out of the system and defining a solid base from which to 
work from.

>So far, gentoo seems to be the distribution I would prefer to use.
>Congratulations on the results you've achieved.

drobbins deserves all the credit.  He's designed and built the only Linux
distro I'll use.

kabau

--
"UNIX was not designed to stop you from doing stupid things, because that
 would also stop you from doing clever things."  --Doug Gwyn

Re: [gentoo-dev] gentoo's portage

[Grant Goodyear]

> 1) documentation:  please ensure the 'build' options
> (/etc/make.conf.build) are explained as to what all the three letter
> acronyms mean and the impact/dependencies of picking them

Probably true, although /etc/make.conf does say pretty clearly not
to use those flags:

# Other build related use flags (don't use these)
# -----------------------------
# pic static build simpleinit 


> 2)documentation:  the departure from 'normal' system startup scripts
> needs to described.  What will happen when a 'normal' non-gentoo
> application tries to install itself expecting a 'standard' layout?

True, we do need much better documentation on our new start-up scripts.
I assume that there will eventually be an IBM developerworks article
on how they work, but we really should have something now.

> 3) portage:  a great drawback seems that 'constant human effort' is
> required to ensure the portage tree is updated.  I can see this becoming
> an issue for gentoo.  There are many applications released and updated
> that must conform to the portage system to make sure 'the latest and
> greatest' really IS available "as advertised" on the web site :) eg.  I
> needed the latest Matrox drivers and Powerdesk utility which I installed
> manually.  my system's portage is not aware of manual installs I perform
> and I needed the drivers before they could be added to the 'gentoo
> portage' tree.  

So far our solution has been to make the ebuild so simple to write 
that it is nearly as fast to write the ebuild and use that as it
is to do the usual ./configure && make && make install series.
There are of course special cases that are not as simple, but often
updating an ebuild for a newer version of a package is as simple
as just changing the name of the ebuild to reflect the new version
number.  If you do write your own ebuilds, please send a copy to
us at gentoo-ebuild.gentoo.org!


> 4) portage: sounds powerful - perhaps it can also be used as a means of
> a useful/automated/'user-friendly' 'uninstall' which is sorely lacking
> in the unix world.  I have tried 4+ distributions of Linux and am
> getting a bit tired of telling their respective 'uninstall' that when I
> want to remove a game, I would STILL like to keep gnome on my system.

Portage uninstalls work very well, but there are a couple of glitches.
It is still possible to have a (broken) ebuild that installs files
directly to the filesystem (instead of using a "fake install" first),
and those files aren't removed by Portage.  Also, right now if gnome
is removed by Portage, the packages that depend on gnome will not be,
thereby potentially breaking lots of stuff.  These issues are being
worked upon, but we're a pretty busy group!

> 5) portage: if possible (space requirements, etc), 'generic 386'
> binaries for the base system should be in the portage tree so that
> initial installs/builds can progress faster.  the 'emerge system' on my
> P4 1.4 took a looong time.  Would be better to have as much generic base
> system and then be able to fine tune as desired.

When 1.0 comes out there will be .iso's with binary packages that
can be installed directly.  It's just that Gentoo development has
been too rapid and unstable recently to make that feasible at the
moment.  I think that Gentoo is stable enough now, however, that
our next release really should have binaries available.

> 6) portage: organization.  consider reducing the # of categories.
> People can mainly remember 7 +/- 2 items of information.  go into
> /usr/portage and there are a lot of choices which will only increase in
> the future.  Consider:

It may be that we just need a better index (or search system), rather
than fewer categories, but I'm pretty agnostic on this one.

Welcome to Gentoo Linux!

-Grant-
-- 
___________________________________________________________________
|     Grant Goodyear                  |  The Secrets of Physics:   |
|     Dept. of Chemistry - Clemson U  |1. Add zero.                |
|     Clemson, SC  29634              |2. Multiply by one.         |
|-------------------------------------|3. Expand in a Taylor series|
|e-mail: goodyea@clemson.edu          |4. Integrate by parts.      |
|www:bernacchi.chem.uh.edu/~grant     |5. Fourier transform.       |
|                                     |6. Add auxiliary variables  |
|_____________________________________|____________________________|
 

Re: [gentoo-dev] gentoo's portage

[Grant Goodyear]

> I think the more you play with the system, the more you'll begin to see 
> it's logic.  Also, I know that some of us would like to see a search 
> feature.  However, that is low on the list for now.  We're in the process 
> of shaking bugs out of the system and defining a solid base from which to 
> work from.

There is a search function, although it's fairly simple:
/usr/lib/portage/bin/pkgsearch  (There's even a pkgsearch man
page.)

-g2boojum-
-- 
___________________________________________________________________
|     Grant Goodyear                  |  The Secrets of Physics:   |
|     Dept. of Chemistry - Clemson U  |1. Add zero.                |
|     Clemson, SC  29634              |2. Multiply by one.         |
|-------------------------------------|3. Expand in a Taylor series|
|e-mail: goodyea@clemson.edu          |4. Integrate by parts.      |
|www:bernacchi.chem.uh.edu/~grant     |5. Fourier transform.       |
|                                     |6. Add auxiliary variables  |
|_____________________________________|____________________________|
 

Re: [gentoo-dev] gentoo's portage

[Nathaniel Grady]

Amen to that! 3 cheers for drobbins!

--nate grady

> drobbins deserves all the credit.  He's designed and built the only Linux
> distro I'll use.
> 
> kabau
> 
> --
> "UNIX was not designed to stop you from doing stupid things, because that
>  would also stop you from doing clever things."  --Doug Gwyn
> 
> _______________________________________________
> gentoo-dev mailing list
> gentoo-dev@cvs.gentoo.org
> http://cvs.gentoo.org/mailman/listinfo/gentoo-dev