Kamil Paral wrote:
I very much agree with this. The official compose gets done just
once,
lowest priority (Note that most tested configurations actually decrease
the compose time, or increase slightly, but nothing major). The image size
is important, so that stuff fits on a flash drive. But the proposed
changes differ in just a few percent increase or decrease. That is not a
major change in any way. On the other hand, the install time differences
*are* major - we can have 30-40% reduction in install time. You can argue
that install gets done just once and is also not an important factor. That
is true, for a standard user. But there are areas where hundreds of
installation are performed each day, and such a big reduction really
matters. Of course, my job, QA, is one of those areas. CI is another. Any
third party vendor-related testing can be impacted in the same way. The
more we steer towards automated testing and continuous integration, the
more important the compose time and installation time will be. And because
usually you do many installations from a single compose, the installation
time is the more important one.
The areas where hundreds of installations are performed from an image are
hardly the common case, and those installations should really consider
composing their own images with exactly the package set that they need
(including site packages that Fedora cannot ship), images which they can
then compress with a faster-to-decompress algorithm. I think that we should
optimize for the standard user who installs exactly once.
And on a slow enough connection (e.g., dial-up, which is still common in
large parts of the world), "a few percent increase or decrease" in download
time can mean hours of difference, much more than even 30-40% of install
time. (Assuming that your numbers are even accurate, which I have not seen
any proof of so far.) Please do not focus on US and European users only.
(And even here in Europe, not everyone has broadband.)
And finally, those few percent can make the difference between fitting on a
given fixed-size physical media or not. The original change proposal of
trying to minimize the size might actually make at least the smaller spins
fit on a DVD again. (Even though compression alone is not a silver bullet to
compensate all the other bloat that has been added over time. It can only
help alleviate the impact.)
Kevin Kofler