On 20 April 2017 at 14:23, Josh Boyer <jwboyer(a)fedoraproject.org> wrote:
On Thu, Apr 20, 2017 at 2:17 PM, Bill Nottingham
> Josh Boyer (jwboyer(a)fedoraproject.org) said:
>> > We are not, though. Modular Fedora, Atomic Fedora, and regular Fedora
>> > are all distinctly different platforms with different delivery
>> > mechanisms and core technologies. Unless you plan to take an axe to
>> > Modular and Atomic Fedora, we're providing multiple platforms.
>> Those are not platforms. Those are ways we compose Fedora, or
>> artifacts of our release. The platform being defined is the set of
>> services and APIs we provide to other things to consume. That is
>> distinctly different than the artifacts they may choose to use.
>> - A time synchronization service and API is defined in the platform
>> - An implementation of that might be ntpd. Or chrony. Or some
>> systemd thing. As long as the API and service remains consistent, the
>> platform is consistent.
>> - Modularity is a mechanism to define these services an APIs at a
>> higher level than per package. It lets us set the platform at "we
>> provide a webserver", not "we provide apache and nginx and lighttpd.
>> take your pick". You can still choose specific webservers, but the
>> module definition for each will hopefully fulfill the platform API.
>> That's one of the goals.
> I think that might be the issue here - it's likely to be seen by most as
> a change in how we describe the platform that is delivered. For better or worse,
> even in the Atomic & Workstation & Spin & so on days, the Fedora
> is likely seen as "a collection of packages, including three web servers,
> five desktops, and as many as twenty IRC clients".
Yes, it's a change. Yes, "platform" is a very overloaded word in the
industry in general. There's no better word though.
The issue is that you need to clearly define these terms before you
can have the conversation... and make sure that people are all on
board with how you see the difference between platform and platforms.
Because if next week Workstation, Server, ostree or arm use the word
platform in the old version.. you are back to square one.
If you wrote code and started using a variable without properly giving
it a zero.. and it works for your initial test you are going to be in
a world of hurt when running it outside of that machine. Different
compilers or architectures may not zero out variables when they are
allocated. You may end up with whatever the mmo last had in its grasp.
People aren't much different. We need to have zeroed out variables
when we are trying to get new distinct definitions even if it take up
a ton of extra instructions and clock cycles because wetware is crap.
> I know the rings -> modularity -> ??? discussions are about
> idea/perception, but as we've still only ever produced the same set of
> artifacts ('a big repo turned into images and isos and variant repos'),
> I don't know that the public perception has changed. We need to get everyone
> on the same page as to whether it's a bikeshed or woodshed before we can
> talk about what color it is.
Modularity has been a Council Initiative for a long time now, with
literal modules being created today and a Fedora Server being composed
from them (Boltron). The rings/core discussions came before that and
are a direct ancestor. This is becoming very real, very quickly. If
it takes changing the Fedora mission statement for people to start
paying attention to the direction we've been headed, that's kind of
unfortunate but also I'll take it. Any crack in the door to get that
messaging and discussion started.
council-discuss mailing list -- council-discuss(a)lists.fedoraproject.org
To unsubscribe send an email to council-discuss-leave(a)lists.fedoraproject.org
Stephen J Smoogen.