Re: removing tophat from Debian
On 23.12.17 09:06, Afif Elghraoui wrote:
على الخميس 21 كانون الأول 2017 08:10، كتب Steffen Möller:
So, what could justify a deprecated piece of software in Debian. To mind
* an API change
* a well-established tutorial that was not yet updated
These two points can be summarized as inertia. I really try to avoid
accomodating that situation by keeping around deprecated
software--either I'd update the reverse-dependency to use the current
CLI/API myself, or leave it to upstream, otherwise keep it out of the
* historic landmarks achieved by that software that one wants to
compare new developments against
So should we also package something like the old Celera assembler
(wgs-assembler) despite it's being unmaintained and obsolete?
I wouldn't consider this reliable because, as the tophat developer
indicated, the program is still widely used despite being obsolete. I
think this is also because of inertia.
Still. Somebody running a larger project over many years may decide to
use the older inferior version to allow comparisons with earlier data.
Of course these folks should upgrade their technology. And they will.
But we should not dictate that transition or know better from remote
what is best for them. And projects running over 5+ years are rather
common. Just think about rare diseases that do not see patients too
often and the not unlikely loss of raw sequencing data - you will treat
new data with an old protocol.
I have difficulties with the role of our distribution as some kind of
package-value police. My personal
threshold for sponsoring is a peer-reviewed publication.
Here, you're already setting a package-value criterion.
But it is a personal one. I do not upload what is not published and I do
not sponsor what is not published. But I am not the only developer /
sponsor, so I do not make any decision for our distribution.
But I like the
role of our distribution as a communicator.
And we should indeed collect ideas on how to spread the news on a better
alternative - any clear cut
improvals are rare enough, though.
If upstream explicitly asks for a removal from our distribution, then I
am still hesitating. Do they also
retract their publication? Most likely not since at some point in
history that previous version was just
fine. It is a document that we do not want to lose. The same for the
software of that time. And I argue
that I do not want to lose the package, either. I am afraid that when we
have it in snapshot.d.o, one or two
releases down the road, the package will be forgotten. We would need
something in our task pages,
then, to point to it, I suggest.
I think this is a good point, and valid for any Debian package. It's not
really trivial (at least as far as I know) to find out whether a package
used to be in Debian. I've always done it by URL hacking, trying
packages.debian.org/<guessed-package-name> and finding the package
history and finally the RoM bug report.
In any case, as far as tophat goes, I think the only reason we have a
problem here is because upstream renamed the program. If hisat is the
continuation of tophat and they just made tophat 3.x and 4.x, we'd right
now have the latest in the archive and the older versions would be in
snapshot.debian.org. This is a similar situation for wgs-assembler and
it's successor, canu.
I am with you, except for the consequence of a removal. There should
be something else. We just do not know what it is.