Re: Package names don't matter too much [was ITP: cddb.bundle -- CDDB Bundle for GNUstep]
On Thursday 07 October 2004 11:46, Frank Küster wrote:
> Tilo Schwarz <list@tilo-schwarz.de> wrote:
> > On Thursday 07 October 2004 10:09, Frank Küster wrote:
> >> Tilo Schwarz <list@tilo-schwarz.de> wrote:
> >> > On Monday 04 October 2004 20:23, Frank Küster wrote:
> >> >> I must have missed this thread... What is ".bundle" meant to
> >> >> indicate?
> >> >
> >> > On NeXT-Step systems the ".bundle" suffix of a directory
> >> > indicates a dynamically linkable module (basically like a shared
> >> > lib).
> >>
> >> Is Debian a NeXT-Step system? I guess not. Does ".bundle" have any
> >> meaning to our *users*? I guess no.
> >>
> >> It doesn't really matter whether you can teach the readers of
> >> ITP's on debian-devel what ".bundle" means. The question is
> >> whether we serve our users with what I would call a "cryptic name"
> >> for at least 90% of them.
> >
> > The intention of my answer was not to argue concerning the gnustep
> > naming issue in any direction - I couldn't care less. The answer
> > doesn't contain a single word regarding the naming discussion.
> >
> > The intention was to answer what I thought was a question of you:
> >> >> I must have missed this thread... What is ".bundle" meant to
> >> >> indicate?
>
> The misunderstanding is that, since bundle has a meaning for
> developers (in a GNUstep context), you read it as 'What is a
> ".bundle"'. I must admit that I wasn't really interested in the
> answer.
Now I know ;-)
> What I meant was: 'What is the message you want to send to the user
> by naming a package "foo.bundle" instead of "foo"'. It seems to me
> the answer is: "There is no message". So my comment is: Why not keep
> it out of the name (after choosing a sensible name for the
> application in the first place)?
Yes, I understand the argument, but let's try to get the discussion a
bit more general.
As I said, as a happy Debian user I don't care at all what package name
a package has (except, maybe, it shouldn't be too long to type). You
might think, well - strange user - but let me try to explain:
From a techical point of view a package name is a key to identify a
package. This key can be fed to various tools (like dpkg, BTS,
apt-cache, ...) to perform some action. The key must be unique, but
that's about it (besides some rules about the characters to be used, I
guess). So for the about 16185 packages in Sarge it would suffice to
have package names with just three letters (26^3 = 17576). (BTW, there
are more than 1000 packages in Sarge having <= four letters).
Now with the package name all the package (meta-) data can be accessed.
A simple "apt-cache show foo" gives me all kinds of information besides
the important Description:-field which is the semantics associated with
the "key" package name. If I search for a package I use something like
"apt-cache search X terminal emulator" and get a nice output with lines
like (besides others)
aterm - Afterstep XVT - a VT102 emulator for the X window system
lynx - Text-mode WWW Browser
rxvt - VT102 terminal emulator for the X Window System
terminal - Terminal Emulator for GNUstep
wterm - An rxvt based, color xterm replacement
I don't mind what "key" is used, the one line description sais it all.
So for me as a user the package name doesn't matter, the package
description (and the other meta-data) matters.
So this nameing discussion is about what package data should be mangled
into the package name, in this case especially what dependency data.
With the same reasoning I could ask to mangle any other information
from the package data into its name (like section, maintainer,
whatever), because there always someone who likes to see that
information. But there's no need to do that, because there are tools to
extract that information right away. E.g. aptitude gives me the most
important stuff with a single keypress.
To conclude, I'd be still a happy Debian user even if Sarge would be
released with just three letter package names ;-).
Regards,
Tilo
Reply to: