[gentoo-dev] new portage categories

[gentoo-dev] new portage categories

[Jonas Bernoulli]

Hello

Recently I started to exclude parts of the portage tree for various
reasons. One of them is that I play with the thought of creating my
own minidistro/livecd based on Gentoo. So keep in mind that I don't
think that it is in general useful to trim the portage tree to the
extend that I have.

One of the obvious things one might start out with is to exclude kde
when one prefers gnome or vice versa. I extended this to exclude
packages within on category that fulfill the same purpose. E.g. I use
app-admin/metalog so it doesn't make much sense to include all the
other loggers. Since the decision for metalog at the time was not
substantiated at the time a also made a note basically saying "is
metalog really the best? these are the alternatives". Once I have the
time I might include the other loggers again and evaluate them.

But to make the list of loggers I basically had to read through the
descriptions of all packages in app-admin to sort out all loggers (I
could have searched but I wanted to make sure I don't miss anything).
Since I have not done this for loggers only this was/is a lot of work
and I therefor asked myself why are there no categories like app-log
or app-cron.

So I ask you: why are there no such categories? Of course I can
imagine a few reasons myself for not having more categories:

(1) a category must in general include n packages
(2) more categories are evil, once we start creating new once there is no end
(3) moving packages in the tree is bad, things break
(4) who does all the work
(5) subcategories would be better, but to implement this...
....

Please point me to any discussions on this subject. Keep in mind I am
not demanding new categories, I am not even asking for them to be
created. I simply would like to know why there aren't more. And if you
developers are also interested in more categories I would love to make
some suggestions and help with looking through the tree to see which
packages would have to be moved.

Reasons why more categories might be usefull:

(1) Easier to find new packages

One great benefit of going through the tree to exclude packages was
for me that I came across many great packages I did not know about.
The likeliness of this would be increased e.g. for the category
app-admin, if I did not have to read the DESCRIPTION of  ~8 loggers
even though I already have selected one and currently am not at all
interested in evaluating the alternatives. In addition there are ~14
logfile analysers of some sort in app-admin ~4 logfile rotators ~2
other log related packages. And some more can probably be found in
app-misc, x11-apps and x11-misc. So all in all at least ~30 packages
that have to do with logging, why not create app-log?

(2) Easier to find alternatives to a package

Need a logger? See what is in app-log!

(3) Makes it less likely that similar packages end up in different categories

Just an example app-admin/pwcrypt DESCRIPTION="An improved version of
cli-crypt (encrypts data sent to it from the cli) but cli-crypt is in
app-crypt"

By the way why is there app-crypt but not app-log?

There are other things I noticed when weeding through the tree. E.g.
some DESCRIPTIONs start with a capitalized letter others don't. Some
end with a period, others don't. Some for now apparent reason start
with "foo is an application to do bar" instead of "do bar".

I understand that it is not very interesting for any developer to fix
such minor errors, and I am not asking anyone to do something about
it. But I would like to know if there is any change that new
categories are created if I or others collect lists of packages that
could be moved to new categories. And yes I understand that the work
doesn't end here and would possibly also help finding packages whose
dependencies would have to be modified.

For me also this is not exactly fun. But since I do this kind of work
for my own personal benefit at the moment I might as well do it in a
way that benefits others as well.

-- Jonas
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] new portage categories

[Donnie Berkholz]

On 18:35 Mon 04 Feb     , Jonas Bernoulli wrote:
> So I ask you: why are there no such categories? Of course I can
> imagine a few reasons myself for not having more categories:
> 
> (1) a category must in general include n packages
> (2) more categories are evil, once we start creating new once there is no end
> (3) moving packages in the tree is bad, things break
> (4) who does all the work
> (5) subcategories would be better, but to implement this...
> ....
> 
> Please point me to any discussions on this subject. Keep in mind I am
> not demanding new categories, I am not even asking for them to be
> created. I simply would like to know why there aren't more. And if you
> developers are also interested in more categories I would love to make
> some suggestions and help with looking through the tree to see which
> packages would have to be moved.
> 
> Reasons why more categories might be usefull:

Sounds like what you really want are tags, not categories ...

You could play with adding them into metadata.xml and patching some 
existing search tools to search for them. I'm all for the idea of tags, 
and I think it's a better approach than categories.

Thanks,
Donnie
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] new portage categories

[Jonas Bernoulli]

On 2/4/08, Donnie Berkholz <dberkholz@gentoo.org> wrote:

> Sounds like what you really want are tags, not categories ...

Yes and no. tags would definitely be better than subcategories. But
for some packages a new category would probably still make sense like
app-scm ( http://www.mail-archive.com/gentoo-dev@lists.gentoo.org/msg27404.html).

> You could play with adding them into metadata.xml and patching some
> existing search tools to search for them.

Added to my TODO list. Extending existing search tool would only be a
first step however. The ability to exclude/include packages from the
tree would also be a useful feature (and that doesn't seam to fit the
design of the tools I have used).

> I'm all for the idea of tags,
> and I think it's a better approach than categories.

Thinking about it again I would say tags and categories just fulfill
different purposes. Tags can not replace categories but might be a
useful extension to categories for the tasks I described, not more not
less. They are not better or worse, just different:)

-- Jonas
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] new portage categories

[Donnie Berkholz]

On 20:11 Mon 04 Feb     , Jonas Bernoulli wrote:
> Thinking about it again I would say tags and categories just fulfill
> different purposes. Tags can not replace categories but might be a
> useful extension to categories for the tasks I described, not more not
> less. They are not better or worse, just different:)

Why don't you think they can replace categories?

Thanks,
Donnie
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] new portage categories

[Vlastimil Babka]

[-- Attachment #1: Type: text/plain, Size: 992 bytes --]

Donnie Berkholz wrote:
> On 20:11 Mon 04 Feb     , Jonas Bernoulli wrote:
>> Thinking about it again I would say tags and categories just fulfill
>> different purposes. Tags can not replace categories but might be a
>> useful extension to categories for the tasks I described, not more not
>> less. They are not better or worse, just different:)
> 
> Why don't you think they can replace categories?

For example:
# eix -e fuse
* app-emulation/fuse
      Description:         Free Unix Spectrum Emulator by Philip Kendall

* dev-java/fuse [1]
      Description:         Fuse is a lightweight resource injection 
library specifically designed for GUI programming.

* sys-fs/fuse
      Description:         An interface for filesystems implemented in 
userspace.

Also imagine all those packages sorted into subdirs just by first 
character of the name (because having them all in one huge dir would be 
murder), yuck :)

-- 
Vlastimil Babka (Caster)
Gentoo/Java


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-dev] new portage categories

[Jonas Bernoulli]

On 2/4/08, Donnie Berkholz <dberkholz@gentoo.org> wrote:
> On 20:11 Mon 04 Feb     , Jonas Bernoulli wrote:
> > Thinking about it again I would say tags and categories just fulfill
> > different purposes. Tags can not replace categories but might be a
> > useful extension to categories for the tasks I described, not more not
> > less. They are not better or worse, just different:)
>
> Why don't you think they can replace categories?

Quick answer: Because there are packages with the same name in
different categories. How would tags deal with that?

Long answer: Well maybe there is a way. But I think that it would
probably take a long time to make such a change. Technically tags
could probably replace categories but then their would be no definite
"full" name for that package anymore.

Someone calls it foo/app and someone bar/app, and since there is also
fuu/app which is a different application but with the same name,
nobody would no for sure about package is being taked about without
checking if his beloved foo/app is the same as bar/app the other guy
is talking about. Also how do you sort the packages in the tree? all
in one directory?! every package in the directories of each tag it
belongs to using hardlinks?

-- Jonas
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] new portage categories

[Alec Warner]

On 2/4/08, Jonas Bernoulli <jonas@bernoulli.cc> wrote:
> On 2/4/08, Donnie Berkholz <dberkholz@gentoo.org> wrote:
> > On 20:11 Mon 04 Feb     , Jonas Bernoulli wrote:
> > > Thinking about it again I would say tags and categories just fulfill
> > > different purposes. Tags can not replace categories but might be a
> > > useful extension to categories for the tasks I described, not more not
> > > less. They are not better or worse, just different:)
> >
> > Why don't you think they can replace categories?
>
> Quick answer: Because there are packages with the same name in
> different categories. How would tags deal with that?

Techincally you could enforce UNIQUE(pkg,[tags]), I agree thats a poor
constraint though ;)

>
> Long answer: Well maybe there is a way. But I think that it would
> probably take a long time to make such a change. Technically tags
> could probably replace categories but then their would be no definite
> "full" name for that package anymore.
>
> Someone calls it foo/app and someone bar/app, and since there is also
> fuu/app which is a different application but with the same name,
> nobody would no for sure about package is being taked about without
> checking if his beloved foo/app is the same as bar/app the other guy
> is talking about. Also how do you sort the packages in the tree? all
> in one directory?! every package in the directories of each tag it
> belongs to using hardlinks?

Filtering packages on tags would be very difficult without a metadata
cache for metadata XML as you would need to pull down each package's
metadata.xml to determine if it meets your filtering requirements.
Excluding individual packages wouldn't be that hard although I imagine
there may be problems with the length of the rsync command line if
your filtering is too aggressive ;)

>
> -- Jonas
> --
> gentoo-dev@lists.gentoo.org mailing list
>
>
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] new portage categories

[Donnie Berkholz]

On 15:12 Wed 06 Feb     , Alec Warner wrote:
> On 2/4/08, Jonas Bernoulli <jonas@bernoulli.cc> wrote:
> > On 2/4/08, Donnie Berkholz <dberkholz@gentoo.org> wrote:
> > > On 20:11 Mon 04 Feb     , Jonas Bernoulli wrote:
> > > > Thinking about it again I would say tags and categories just fulfill
> > > > different purposes. Tags can not replace categories but might be a
> > > > useful extension to categories for the tasks I described, not more not
> > > > less. They are not better or worse, just different:)
> > >
> > > Why don't you think they can replace categories?
> >
> > Quick answer: Because there are packages with the same name in
> > different categories. How would tags deal with that?
> 
> Techincally you could enforce UNIQUE(pkg,[tags]), I agree thats a poor
> constraint though ;)

I would probably print the ambiguous package name, with info on each 
package, and a list of tags unique to each that could be used to specify 
which one you want. Perhaps a numbered list too.

Thanks,
Donnie
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] new portage categories

[Marius Mauch]

On Thu, 7 Feb 2008 12:56:43 -0800
Donnie Berkholz <dberkholz@gentoo.org> wrote:

> On 15:12 Wed 06 Feb     , Alec Warner wrote:
> > On 2/4/08, Jonas Bernoulli <jonas@bernoulli.cc> wrote:
> > > On 2/4/08, Donnie Berkholz <dberkholz@gentoo.org> wrote:
> > > > On 20:11 Mon 04 Feb     , Jonas Bernoulli wrote:
> > > > > Thinking about it again I would say tags and categories just
> > > > > fulfill different purposes. Tags can not replace categories
> > > > > but might be a useful extension to categories for the tasks I
> > > > > described, not more not less. They are not better or worse,
> > > > > just different:)
> > > >
> > > > Why don't you think they can replace categories?
> > >
> > > Quick answer: Because there are packages with the same name in
> > > different categories. How would tags deal with that?
> > 
> > Techincally you could enforce UNIQUE(pkg,[tags]), I agree thats a
> > poor constraint though ;)
> 
> I would probably print the ambiguous package name, with info on each 
> package, and a list of tags unique to each that could be used to
> specify which one you want. Perhaps a numbered list too.

That only works for the user interface, not so good for depend strings,
config files and pretty much everything else.

Marius
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] category ambiguity

[Donnie Berkholz]

On 00:12 Fri 08 Feb     , Marius Mauch wrote:
> On Thu, 7 Feb 2008 12:56:43 -0800
> Donnie Berkholz <dberkholz@gentoo.org> wrote:
> 
> > On 15:12 Wed 06 Feb     , Alec Warner wrote:
> > > On 2/4/08, Jonas Bernoulli <jonas@bernoulli.cc> wrote:
> > > > On 2/4/08, Donnie Berkholz <dberkholz@gentoo.org> wrote:
> > > > > On 20:11 Mon 04 Feb     , Jonas Bernoulli wrote:
> > > > > > Thinking about it again I would say tags and categories just
> > > > > > fulfill different purposes. Tags can not replace categories
> > > > > > but might be a useful extension to categories for the tasks I
> > > > > > described, not more not less. They are not better or worse,
> > > > > > just different:)
> > > > >
> > > > > Why don't you think they can replace categories?
> > > >
> > > > Quick answer: Because there are packages with the same name in
> > > > different categories. How would tags deal with that?
> > > 
> > > Techincally you could enforce UNIQUE(pkg,[tags]), I agree thats a
> > > poor constraint though ;)
> > 
> > I would probably print the ambiguous package name, with info on each 
> > package, and a list of tags unique to each that could be used to
> > specify which one you want. Perhaps a numbered list too.
> 
> That only works for the user interface, not so good for depend strings,
> config files and pretty much everything else.

I don't think categories are the best way of resolving ambiguities, 
because they don't uniquely identify a package. One could imagine two 
packages in the same category with the same name (for example, two 
Python modules that do drastically different things but would go in 
dev-python).

I'm not sure what the best way is, but I don't think it's categories. 
Perhaps some sort of UUID for packages? You could treat unique tags like 
categories, but those could also be duplicated.

Thanks,
Donnie
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] category ambiguity

[Nirbheek Chauhan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Feb 8, 2008 7:32 AM, Donnie Berkholz  wrote:
> I don't think categories are the best way of resolving ambiguities,
> because they don't uniquely identify a package. One could imagine two
> packages in the same category with the same name (for example, two
> Python modules that do drastically different things but would go in
> dev-python).

In my humble opinion, the current classification is perfectly fine in
practice -- it is quite similar to the human naming system:

,  => / (okay, not exactly -- categs are limited -- but close enough)

Surely there are more humans than packages, but we do fine don't we? :)
And wherever there is a chance of conflict, we created aliases.

This system of classification appeals to my common sense. If there are
two packages with the same name in the same "general category", say,
for example, perl/python modules/libraries, there will be confusion
between them in the real world anyway, and one of them will *have* to
change their name to prevent it. Off my head, I can think of pkgconfig
crashing and burning if it got involved in such a dispute :^)

>
> I'm not sure what the best way is, but I don't think it's categories.
> Perhaps some sort of UUID for packages? You could treat unique tags like
> categories, but those could also be duplicated.

I really don't think this needs such complicated solutions; especially
when the "problem" is actually more of an enhancement to the category
system, and the "solution" will complicate more than it will
alleviate.

Let's not forget that most package/dependency management systems
(pkgconfig, PackageKit[?], dpkg, rpm, etc) make do with just one
name-space -- staying one step ahead should be enough ;)


- --
~Nirbheek Chauhan

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)
Comment: http://firegpg.tuxfamily.org

iD8DBQFHrCWZb1z91vbKYbYRAq7AAJ96IukPElYbcamfct32NLbsRCZUjQCgxh1X
JonN+76MwRb5D06b8ZWghjE=
=TwMR
-----END PGP SIGNATURE-----
-- 
gentoo-dev@lists.gentoo.org mailing list

Re: [gentoo-dev] category ambiguity

[Nirbheek Chauhan]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Feb 8, 2008 3:19 PM, Nirbheek Chauhan  wrote:
> In my humble opinion, the current classification is perfectly fine in
> practice -- it is quite similar to the human naming system:
>
> ,  => / (okay, not exactly -- categs are limited -- but close enough)

[last name], [first name] => [categ]/[app] (okay, not exactly --
categs are limited -- but close enough)

Hmm, strange, I could've sworn I was using plain-text formatting



- --
~Nirbheek Chauhan

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)
Comment: http://firegpg.tuxfamily.org

iD8DBQFHrCYJb1z91vbKYbYRAjy/AJ4so+lfU9MOYBNNKF3MU9bTc2BxTgCfWfT2
x56+copSI//8KPlrFuNgNJI=
=sQEi
-----END PGP SIGNATURE-----
-- 
gentoo-dev@lists.gentoo.org mailing list