Gary Buhrmaster wrote:
Some people download once, and install once. For
those, it is pay me now, or pay me later, and it may
be a wash, time wise, depending on your download
speeds, but it is also just a one time thing in any
case. Some people download once, and install lots
of times (in which case the install time may dominate).
There is no one size fits all.
FD: I am in the later group (download once, install on
many VMs/systems during the lifetime of a release),
so would likely prefer to see something like zstd gain
traction at some point, but that is a different discussion.
There is also another group of VM users (mainly testers), who never actually
install the ISO, but only run it in live mode, in which case it is only the
live system boot time and runtime performance that are relevant, not the
installation time. Those metrics are also affected by the decompression
speed, but not necessarily to the same extent.
(This is not how the Fedora QA team's testers usually work because the
Fedora QA test protocols include installation, because testing the installer
and the installed system is an essential part of distribution QA. But not
all users doing live image testing in VMs are going to test the installer
this way. We should be careful not to assume that the usage patterns in
Fedora QA are relevant to all end users and optimize exclusively for those
usage patterns.)
Kevin Kofler