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

Re: tclap-1.1.0 comments



Hi Mike,

On Tuesday 20 November 2007 07:15:12 Andreas Tille wrote:
> On Mon, 19 Nov 2007, Dirk Eddelbuettel wrote:
> > On 19 November 2007 at 17:06, Mike Smoot wrote:
> > | I work on another free software project (http://cytoscape.org) and one
> > | of my collaborators is trying to put together a Debian package for
> > | Cytoscape, but is having some difficulty.  I don't know anything about
> > | the Debian packaging, so I was wondering if I could put him in touch
> > | with you to possibly get a few pointers?  If it's ok with you, I'll
> > | copy you both from my work mail account.  Of course, if you're too busy
> > | I also completely understand!
> >
> > I am unfortunately super-busy, but as you have always been so helpful
> > with tclap, how could I say no?  :-)
> >
> > That said, cytoscape looks like a rather neat science / med project we
> > should try to get into Debian. I am cc'ing Andreas who is the goto-guy
> > for 'Debian Med'.  Andreas, could you help as well?
>
> Many thanks for the hint to this interesting project which is definitely
> interesting for Debian-Med.  My suggestion would be that you (Mike) or your
> collaborator who actually has the problem just posts a status mail which
> mentions the problems on the Debian-Med mailing list.  We will try to help
> out as best as we can.  Perhaps we will need some help from debian-java
> because I'm afraid the experience in Java packaging is not (yet) wide
> spread in the Debian-Med group.  If there is a general interest in Free
> Software in bioinformatics you might consider subscribing to the relatively
> low volume list.

we are working with Cytoscape locally. It is quite a tool that drags quite a 
number of Java libraries with it. To have just something that works on Debian 
as a package you would leave the upstream source as it is, make sure that the 
compilation works, and that is basically it.

That speaking, to have something that is decently behaving there should not be 
any redundancy with current existing Java libraries that are already 
distributed with Debian. You'd hence remove all the jars from the upstream 
distribution and work with those already available via /usr/share/java. The 
problem here is that most Java packages in Debian today are not versioned. 
This is a bug as I would say. And, which is even worse, one rarely exactly 
knows about the exact version of  a Java package that a suite is using and 
where to get the source of it from. In Debian, everything that is on the hard 
disk shall be recreateable from the source. The use of maven (which is 
available in Debian now) and the inclusion of online repositories for the 
build process has not really helped this issue.

I am not certain if the distribution of a Java package that contains jars 
without access to their source would be OK to be distributed in the non-free 
section of Debian. Possibly so. For Java, the disadvantages of the non-free 
sections are less severe since there are no other platforms that the tool 
would need to be auto-compiled on. I personally see the main gain of a 
complete Debianisation though in the complete transparency of the inner 
functioning of the tool at hand.

http://pkg-escience.alioth.debian.org is an attempt to gather libraries for 
Java aiming at a package for MyGrid's Taverna. You might find some useful 
bits in there, the effort has stalled, though.

The other issue that comes with Cytoscape is the handling of databases, 
pathway databases in this case. 

Best,

Steffen








Attachment: signature.asc
Description: This is a digitally signed message part.


Reply to: