Poppler packaging suggestion

[dg1727]

Hello, 

About Poppler, I have a packaging suggestion which may affect 
Ubuntu installer code as well.  

First some background:  

In order to render CJK fonts in PDF files, Poppler needs the 
package "poppler-data" to be installed.  If "poppler-data" isn't 
installed, any CJK text in a document is invisible.  

It is stated in https://bugs.launchpad.net/ubuntu/+source/language-
selector/+bug/893920 that "poppler-data" shouldn't be a regular 
(mandatory) dependency because "poppler-data" is too big to fit on 
an Ubuntu install CD.  

An intended solution has been made (documented in 
https://bugs.launchpad.net/ubuntu/+source/language-
selector/+bug/893920 ) by modifying "language-selector", starting 
with Ubuntu 12.04 LTS, to install poppler-data unconditionally.  

There are some cases in which I think this intended solution might 
not work:  

 A. I don't know much about language-selector, so I suppose there 
may be cases when language-selector doesn't run, either on upgrade 
from a version of Ubuntu prior to 12.04, or on a new install of 
12.04 or higher.  

 B. As I recall, once the OS (12.04 or higher) is installed, a 
notification icon prompts the user to run "language-selector", but 
the user may ignore this notification.  Until language-selector is 
run, poppler-data doesn't get automatically installed.  

 C. Someone may un-install poppler-data, either accidentally or on 
purpose, (if a PDF reader was un-installed as part of this, then 
the same or a different PDF reader may then be re-installed without 
poppler-data,) then try to view a PDF file that contains CJK fonts. 
 

My idea is ...

... to make "poppler-data" a Recommended dependency of Poppler, and 
that certain use cases will be as follows:  

 1.  CD installs will enable the apt-get flag "--no-install-
recommends", so that APT won't try to install poppler-data from the 
CD, since poppler-data won't be on the CD.  

 2.  The existing installer has a prompt similar to "Do you want to 
update all software to the latest versions via the Internet?".  
When this is selected by the user and packages are being 
downloaded, "--no-install-recommends" can be disabled (possibly 
letting the user override this in case they are very low on disk 
space), so that poppler-data will be downloaded and installed.  

 3.  DVD installs can disable "--no-install-recommends", and have 
poppler-data on the DVD.  Poppler-data will be installed 
unconditionally from the DVD, without the user being prompted, 
without network access being needed, and without requiring the user 
to run "language-selector" from the desktop notification.  

 4.  The "--no-install-recommends" flag can be disabled by default 
in the GUI frontends to APT (it may be already), so that in case C 
above, poppler-data will automatically get re-installed along with 
any PDF reader that uses poppler-data, unless the user has enabled 
this flag.  

As stated in 
https://bugs.launchpad.net/ubuntu/+source/poppler/+bug/197537/commen
ts/7 :
(The bug summary is "Can't read PDF file with CJK 
(Chinese/Japanese/Korean) text".)  
"The fix should be somewhere in the poppler, it's not applicatin 
task to care about cidToUnicode mask." 

... so I am suggesting that poppler-data be made a dependency of 
Poppler, not of a specific PDF-reading application.  

I hope that labeling certain dependencies as Recommended, and 
adjusting the "--no-install-recommends" strategy as suggested, may 
be helpful in cases other than poppler-data, for packages that 
users would generally want to have but that don't fit on an install 
CD.  

The issue of invisible "CJK" fonts can affect PDF documents that 
contain only Latin characters but were produced by software that is 
set up for CJK speakers.  So the issue can affect non-CJK-speaking 
users.  

Thanks in advance for comment.  

-dg1727