Re: Status of dict-freedict-*-* packages
Andreas Tille schrieb am Freitag, den 23. Januar 2004:
> A more detailed research showed up the fact that all these archives
> contain preprocessed dict-files which are "not really" *sources* in
> the sense of the files you edit to fix bugs. So I dived into the
> description part of the german dictionary files which showed the
> following string:
>
> 27. Apr. 2000 Database was converted to TEI format and
> checked into CVS10.April 2000The Data was originally
> generated from Frank Richter's "ding" Database.It was
> converted to TEI format by Horst Eyermann. Phonetics was
> added throughtxt2pho (a TTS front end for the German
> inventories of the MBROLA project)Mr. Portele granted the
> right to distribute the result of the phoneticconversion
> within this project - thanks.
Please keep in mind, that txt2pho
(http://www.ikp.uni-bonn.de/dt/forsch/phonetik/hadifix/HADIFIXforMBROLA.html)
isn't DFSG free. You're allowed to copy it for free, but you only get
sources for this. So the result of txt2pho may be free, but it's not
possible to create this file from trans-de-en with only DFSG free
tools :-(
> Regarding to the ding database: It is inside Debian in the
> trans-de-en package (Maintainer in CC). So my suggestion is: Use
> the trans-de-en package together with throughtxt2pho to build the
> dictd dictionary from *real* source.
I'd like this idea except the above mentioned problem, that txt2pho
isn't DFSG free.
> This would enable us to orphan the other dictd dictionary which is
> based on the same source (Frank Richter's "ding" database) - the
> dict-de-en package. This package has some bugs assigned which also
> have their trouble because of having no real source.
One way would be using dictunformat to "decompile" the dictionary and
after this use dictfmt and dictzip to compile it again. But this
doesn't solve the problem, that we work on a preprocessed source,
which doesn't allow to merge upstream changes into the package.
> I would like to hear your opinion about building one unified
> dict-de-en package from the trans-de-en package and remove
> dict-de-en as well as removing the dict-freedict-ger-eng package
> which would be superflous in this case.
Sounds like the correct solution to me (especially with the
background, that installing dict-de-en causes a segmentation fault in
dictd).
The disadvantage of this way is that it will remove the txt2pho
created phonetics from the dictionary.
> It just build-depends from trans-de-en and builds the dict
> dictionary from the installed database. On the other hand I see no
> reason why this should not be done as multi-binary package from the
> trans-de-en source.
Me too, except the txt2pho part...
Except this we should think about the encoding of the dictionary. On
my iso-8859-1 system, wordinspect (a dict client) doesn't find any of
the utf-8 encoded non-ascii words from dict-freedict-deu-eng, while I
read in bug #217210 that the iso-8859-1 encoding of dict-de-en is a
problem of utf-8 users...
Tschoeeee
Roland
--
* roland@spinnaker.de * http://www.spinnaker.de/ *
Reply to: