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

Re: texlive-basic_2005-1_i386.changes REJECTED

On Tue, 29 Nov 2005 00:40, Thiemo Seufer wrote:
> Andrew Vaughan wrote:
> > On Mon, 28 Nov 2005 22:28, Thiemo Seufer wrote:
> > > FWIW, Debian package names prefer e.g. foo-en-uk-doc over
> > > foo-documentation-ukenglish. 
Can you provide a reference/stats to back this up. 

(on sarge) 
$ apt-cache search doc |grep -e'-doc-[a-z][a-z] ' |wc -l
$ apt-cache search doc |grep -e'-[a-z][a-z]-doc ' |wc -l

Examination of the first set shows 1 false positive:
gmt-doc-ps - PostScript docs for the Generic Mapping Tools
The other 11 translated docs
Examination of the second set shows:
adduser-ng-doc - Documentation for AddUser-NG users
gnome-db-doc - frontend to the GDA architecture for GNOME -- documentation 
gri-ps-doc - PostScript manual for gri, a language for scientific graphics.
libinti-gl-doc - GtkGLExt bindings for Inti - shared libraries
proj-ps-doc - PostScript docs for cartographic projection filters and library

Also look at the output to apt-cache search firefox |grep locale

> > > This allows to filter documentation 
> > > packages by name (doc-* or *-doc), and following the standardized
> > > ISO abbreviations also seems to be better than using yet another
> > > scheme.
> >
> >
> > To me the hierarchy tree
> > <package>-<sub-package-type>-<language/locale>
> > is much more natural than
> > <package>-<language/locale>-<sub-package-type>
> It may look more natural, but it makes pattern matching harder
> (e.g. python-docutils is a false positive for the naive approach).

Probably any approach will yield false positives with naive search strings.  

However a few false positives don't matter in the typical use case of an 
interactive user searching for a package.
> Thiemo

Package naming should be about what makes most sense to the users.  For me, 
that means increasing specialisation (of package) means tacking additional 
suffixes on the end of the packagename.

package foo.

add a doc package 
foo, foo-doc

add translated docs
foo, foo-doc-en, foo-doc-jp, foo-doc-fr etc.

(Regardless of which way is considered better, this sort of thing should be 
standardised and defined in policy IMO).

Why do you need to filter for *-doc packages anyway?  If you are looking for 
say Japanese doc packages, filtering on say *-doc-jp is as easy as *-jp-doc 
isn't it?  

If you want all doc packages, regardless of language/formatting/encoding, then 
filtering for doc-* and *-doc is not enough anyway.  In order to find 
*-doc-html, *-doc-ps you need you need to add *-doc-* anyway.
  (i) You are going to miss packages which don't contain doc in the name. 
      (eg ada-reference-manual, apt-dpkg-ref, docbook-defguide).
  (ii) In order to find *-doc-html, *-doc-ps etc packages you need to add
       *-doc-* anyway.  (eg exim4-doc-html, exim4-doc-info, r-doc-html,
       r-doc-pdf etc.  Note that renaming these to say exim4-info-doc suggests
       naive users that this is the documentation for the exim4-info package.)
  (iii) So you need to search for doc-* *-doc and *-doc-* anyway.
		 eg. apt-cache search doc | grep -e'^doc-\|-doc-\|-doc '
        or apt-cache search doc | grep -e'^doc-\|-doc\b'

Andrew V.

Reply to: