[prev in list] [next in list] [prev in thread] [next in thread] 

List:       koffice-devel
Subject:    Re: Library for MS Word filters (again)
From:       Werner Trobin <trobin () kde ! org>
Date:       2002-07-19 18:30:45
[Download RAW message or body]

On Thursday 18 July 2002 22:08, Andrew C. Oliver wrote:
> > This is as a follow-up to TJ's recent post in koffice, as I believe it's
> > better to go to koffice-devel first. Feel free to cross-post.
> 
> TJ Brought this to my attention and so I've joined the list.  I'm the 
founder of 
> the POI project.  You can read about me here: trilug.org/~acoliver.  

Hi!
 
> I don't use KDE, but I'm supportive of your efforts and will help any way I 
can
> if you want to use POI.  
> 
> That being said, IANAL (though I do watch law and order on TV sometimes) 
> and am not interested in discussing licensing, Stallman's philosophies or 
any
> of the such, I'm a mere programmer and am only capable of helping in a 
technical 
> capacity.  Should you require assistance with Licensing or other matters 
like that
> please contact the Apache Software Foundation itself as they have people who 
understand
> licenses and other plauges of humanity and law.

Hmm, okay, but who choose the specific license of this project? I really don't 
care about licensing in a political sense, and don't get me started about 
RMS, but nonetheless we have to be very careful with "grey area" licensing 
issues. I don't know if you know the history of KDE, but we had an ugly 
amount of mud throwing just because of licenses.

All I understood is that the (L)GPL and the Apache licenses aren't compatible 
(and I'm not a lawyer either). Due to that it's impossible to use POI for 
KOffice as the licenses are now.
 
> > Although T.J mentions wv2 and POI, I think we should not forget our
> > _current_ MS Word filter. As Shaheed pointed out, it's already in a good
> > shape and only needs improvements to make it someday on par with OO or at
> > least Abiword's wv.
> 
> Cool, we'll take a look at it.  The POI::HDF filter is very early, in the 
end
> I'm sure it will reach the same quality as HSSF, but for the moment its 
pretty new.
> If you have something further along than it, I'd recommend using it and 
consider
> converting later if you desire.

Oh, one question at that point: do you plan to support reading only or do you 
also plan to add writing support?
 
> > Another issue that needs to be examined is embedding support. AFAIK our
> > current filter framework support embedding quite well, i.e Excel inside
> > Word will be KSpread inside KWord, CMIIW. How is the status of such
> > embedding support in POI (HSSF and HDF) ?
> 
> I don't understand how this will be an issue in the end.  POIFS can read the
> various embedded files and pass them to the associated filters.  You can 
choose
> the associated filter (HSSF/HDF) and pass it through.  The "embedded" look 
and feal
> is not the concern of POI.

Right, but sometimes this just isn't enough. Embedded images need information 
from the embedding document (in certain cases), so they somehow have to 
communicate.

> > OTOH, I'm quite glad this topic is heated again :-) To Werner, even though
> > the code wasn't changed in the last month, at least there's little
> > activites (better than none). Like what has been mentioned, it's also
> > quite annoying to improve Excel filters but the certain features in
> > KSpread (such as functions) are missing which should explain why somehow I
> > moved to "fix" KSpread first.
> 
> > I hope I'll be free next month and then able to keep my promise about (yet
> > another) OLE storage library for KOffice. I believe libole2 will die soon,
> > Jody Goldberg (from Gnumeric's fame) develops libgsf as a replacement for
> > libole2 (among others) as he stated in gnome-office mailing list. Gnumeric
> > and Abiword slowly switch to libgsf.
> 
> Eventually I hope to write a C or D (http://www.digitalmars.com/d 
http://opend.org) version of 
> POI.  This will be easy for HDF which is largely generated from XML record 
descriptors, but
> it will be much harder for POIFS due to the inherent complexity of the 
format.
> 
> I highly do not recommend using libole2 - its HORRIBLE code and was 
practically useless when
> writing POIFS.  I'm not sure I recommend writing your own OLE 2 CDF thing, 
POIFS is great if 
> you like (probably the most correct opensource implementation), and 
hopefully libgsf will be
> good, but why start from scratch.  One thing I'd offer on this is that POIFS 
is very well tested
> and has lots of delicious unit tests to make sure it stays that way.  HSSF 
is nearly as tested
> and by the end of the 2.0 release will have full support for graphing and 
formulas.

Sounds cool. This raises the next obstacle, though. I understand that all your 
code is written in Java. To use it in-process we would have to use gcc 3.1 on 
all platforms, which seems to be a really odd restriction.
If we'd use the filter out-of-process then we would have to define some 
communication protocol... hmm, that would at least be inconvenient but 
possible.
 
> Let me know if there is anything I can do to help should you decide to use 
POI for a filter.

I have to admit I just had a very brief look at the POI pages and I somehow 
missed how the content is converted to the "final" format. Did you define 
some intermediate XML format and work with XSLT from there? Does your library 
provide some kind of callback interface?

Thanks,
Werner
_______________________________________________
koffice-devel mailing list
koffice-devel@mail.kde.org
http://mail.kde.org/mailman/listinfo/koffice-devel
[prev in list] [next in list] [prev in thread] [next in thread]