[Q-e-developers] qe package
p.giannozzi at gmail.com
Fri Aug 19 21:56:24 CEST 2016
On Thu, Aug 18, 2016 at 10:52 PM, Lorenzo Paulatto <paulatz at gmail.com>
> If we are really concerned about size, we should just compress the tar
> archive with xz instead of gzip (tar -cJvf ...), which would make it 3 or 4
> times smaller.
3 or 4 times smaller than gzip? maybe 25% smaller (which is not bad at all):
-rw-r--r-- 1 giannozz giannozz 15308784 May 6 17:32 espresso-5.4.0.tar.xz
-rw-r--r-- 1 giannozz giannozz 18305717 May 6 17:32 espresso-5.4.0.tar.bz2
-rw-r--r-- 1 giannozz giannozz 20021535 May 6 17:32 espresso-5.4.0.tar.gz
-rw-r--r-- 1 giannozz giannozz 53903360 Aug 19 19:25 espresso-5.4.0.tar
> Another reason to put more stuff in the main tgz is that some (most?)
> computing centers disallow direct Internet connection from the ssh nodes,
> which causes the auto download to fail
computing centers should solve problems instead of creating them.
On Thu, Aug 18, 2016 at 10:21 PM, Nicola Marzari <nicola.marzari at epfl.ch>
> many times these codes are installed centrally once for all by a sysad,
> and users are discouraged to run/compile independently their own
> executables - a good policy, if you think.
a good policy, as long as the sysadm is more competent and less lazy than
> None suggested an even more baroque set up - my desire was to simplify
> things, and I am convinced a single file or a single script that installs
> everything would benefit everyone. The argument that it costs more to
> download seems just too hard to take
consider that there is a disproportionate number of fake downloads,
presumably by bots.
On 18 Aug 2016, at 11:13, Filippo Spiga <filippo.spiga at quantum-espresso.org>
> [...] I have always believed that the reason we had many packages is to
> avoid a monolithic heavy distribution.
Based on what I see, the core source code is not "that big" in size.
Stressing the "non-monolithic" character of QE is actually the most
important reason behind the current way QE is packaged. Another reason is
that some "general-purpose" packages are vastly more used (and usable) than
more specific ones.
I personally see some beauty and some practicality in changing the
> packaging process.
I am personally happy with the current packaging, but not so happy to rule
out other options, as long as this doesn't interfere with more urgent
things to be done in the next days or weeks in view of the release. I
wouldn't object to include more packages in the "base" tarballs, maybe
stripping out tests and examples, while keeping the possibility to install
on demand, like it is now.
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the developers