Re: edgeR no longer with us - effect on seqcluster unknown

[Andreas Tille <andreas@an3as.eu>]

On Mon, Jun 24, 2019 at 08:47:00PM +0200, Andreas Tille wrote:
> On Mon, Jun 24, 2019 at 07:22:54PM +0200, Steffen Möller wrote:
> > 
> > > > At the very moment I cannot tell about how essential edgeR may be for
> > > > seqcluster and the bcbio workflow as a whole - but Debian not offering
> > > > that package at all - it is a bit weird.
> > > I think what is really weird is the license of locfit (see discussion in
> > > #731599) and the fact that it seems impossible to reach locfit upstream.
> > Agreed.
> 
> > > If you ask me there are chances to talk to CRAN / Bioconductor people
> > > that they are relying on non-free software (which is as far as I know
> > > also against their polica) but nobody has explained that issue clear
> > > enough.  May be its not to hard to create a locfit alternative.
> > 
> > EdgeR allows locfit to be substituted by other estimators.
> 
> At the time when we had the r-bioc-edger package locfit was not
> mandatory.  However, I had the feeling that it became default and thus
> it was not possible to keep it in main.
> 
> > Once I feel
> > a bit more mentally superior to it all, I am tempted to suggest a free
> > alternative
> > to the edgeR developers as a default. But any such scientifically relevant
> > change should come from upstream, not as an unsuspected patch within Debian.
> 
> Fully agreed - so did you contacted upstream?

I guess you will have noticed my mail which also went here to our list.
I think it would not harm to explain CRAN and Bioconductor strictly,
that locfit is not only non-free CRAN is violating the license by not
adding a copy of it.
  
> > Anyway - from what I oversee, the missing edgeR is not critical for bcbio.
> 
> I'd love to have it back anyway.

I droped notes in bcbio/debian/TODO.  Everything listed R package that
does not depend from locfit is now in new. 

I also realised that pytz is packaged and available as python3-tz.  This
leaves python2-dateutils[1].  For me it seems basically unmaintained
upstream and I wonder whether seqcluster is really using it.

Kind regards

     Andreas.


[1] https://pypi.org/project/dateutils

-- 
http://fam-tille.de