Road to Stretch: let's stop increasing major version number in critical libraries at this point
Hi,
I've seen a number of package maintainer willing to upgrade to major
version of packages at this point in time. Among the disruptive changes
that have been done (or are planned to do):
- Django 1.10: uploaded last august, after I wrote more than 30 patches,
there's still issues, it seems. See #843108
- OpenSSL 1.1.x
- SQLAlchemy 1.1: currently in Experimental, this version will break
major OpenStack packages (like Cinder), and testing using the new
version implies a *major* work of full multiple human days at least. I
hope the package maintainer will continue to hold on uploading it to Sid
until after Stretch is released, or until I have finished testing and
fixing. [1]
- JQuery: it was 1.7.2 end of last month, now it's 3.1.x. To me, it is
already very late in the dev cycle of Debian at this point to update. I
don't understand why jquery stayed at version 1.x for so long during the
Stretch cycle, and suddenly so close to the freeze, the package
maintainer decided to upgrade *now*. The first 3.x beta and RCs of
JQuery were published about a year ago, and the 2.x series is even older.
Finally, with the above examples as illustration (and please, these
aren't attacks in any way...), I guess what I'm trying to say here is:
While disruptive changes are necessary evils so we upgrade everything to
the latest version, I would like the Stretch freeze to be as short as
possible. Therefore, I'd prefer if my DD friends were holding on
upgrading to (non bugfix only) new upstream releases of libraries.
Cheers,
Thomas Goirand (zigo)
[1] I'll try to finish (or in fact, restart) the testing of OpenStack
Newton with SQLA 1.1.x, and I very much appreciate the effort to wait
until I'm done with it. FYI, progress is tracked here:
https://etherpad.openstack.org/p/sqla-1.1-transition
Reply to: