On Sat, 2011-07-23 at 00:09 +0000, "Jóhann B. Guðmundsson" wrote:
On 07/22/2011 05:02 PM, Ric Wheeler wrote:
> I don't have it backwards - my day job is to work on exactly getting
> this kind of cutting edge feature out to end users.
Which feature of this magnitute have you put out to your users?
Serously if you have some motivation tips to get things moving I and
others would gladly like to hear them..
> When you propose and implement a massive change, you need to make
> sure that the system as a whole still works.
>
Certanly.
> Of course, the developers of code components *should* work with the
> systemd team, but if they don't (don't like it, have higher
> priorities, whatever), then the proponents of change need to do the
> heavy lifting. Heavy lifting can involve education, convincing
> others in a *nice* way or actually having to do the code.
Regarding your education response we have man pages upstream
documentation Lennart's blog posts and various interviews he has done
along with very active upstream mailings and irc channel and if man
pages or other documentation are *vague* about any issue only thing
people need to do is ask.
You don't receive answers if you don't ask and we can't force
knowledge upon people or have them to educate themselves.
Regarding your convincing people this is the third release cycle that
we have to act nice repeat answers to both end users sysadmins and
developers alike and at the same time be our work ridiculed belittled
and we subdued to personal attacks.
Brace yourself, I think you should take it with philosophy and expect
the same to continue for another 3 releases.
Changes that re this disruptive upset many, take a deep breath and don't
take it personally even when people "attack" you. If you do otherwise
you'll only have a bad life and nothing else.
You may think third time is the charm and I honestly don't how
Lennart
does cope with this but I decided to draw the nice line here the third
release cycle we are put through this.
As much as I would like us to have the authority to "do the code" that
unfortunately wont cut it since we lack the expertise, the experience
and insight that maintainers have through maintaining their component
so having us walsing in and do the conversion might actually be doing
more harm then good however working together will deliver the best
result.
That's where's Ric suggestion comes into place. You have to go a knock
to the door of these developers and have a discussion to let them
understand the importance of the work to be done. It is time consuming
and sometimes awkward but you gotta do it.
> We simply cannot ship something that would lose data or cause
> critical components to fail for our users.
I completly agree and from my pov it's like maintainers do not
understand how important that is for us to manage to convert all the
500 - 600 daemons in one release cycle.
Beside the obvious unfortinate experience our user may go through like
the one Reindl Harald and perhaps others have experienced then this
may also break and hinder work for other developers that build their
application stack on top of those daemons.
One such example is the freeipa stack where the freeipa team needs to
rewrite/replace and or add seperated systemd spesific code and if I'm
not mistaken they will only support sysv on sysv platforms and systemd
on systemd platforms instead of trying to support mixed match of both
init system ( Freeipa guys feel free to correct me if I'm wrong ).
That's the plan supporting a mix seems too difficult.
Freeipa is built on top of 389 Directory Server, MIT Kerberos, NTP,
DNS,and Dogtag ( which I dont recall having any daemons ) of that
only the NTP has been converted work is being done on converting 389
and I think the native systemd units for freeipa server daemons are
ready but they are not being shipped yet however that leaves out bind
and mit kerberos with no movement what so ever on those two components
atleast there has been no respons on the relevant bug report for those
two components as is with so many others and I have not heard a word
from either of those maintainers on irc.
We'll get those components converted in time hopefully we (FreeIPA) are
in close contact with both the MIT Kerberos and Bind maintainers.
What you forgot there is that Dogtag is not daemonless, on the countrary
it depends on Tomcat, which looks like a mess to convert to systemd.
Tomcat seem to be the bit that may stall the whole conversion to be
honest.
The other issue we have is that we need to dynamically generate the list
of service when the first component (dirsrv) starts, as we keep the list
of active services per FreeIPA server in LDAP. But we seem to have a
solution for that, more or less...
Now their new code depends on all of those components to be
converted
and it does not take more then one of those components to not convert
and their whole work is being thrown out the window and they are
forced to rewrite their code to support a mixed batch of sysv legacy
scripts and native systemd units and they have the time from from Beta
to GA to test that code on that mixed batch of init files which at the
same time those unit files might be subdued to various testing and
rewriting as an indvidual standalone services.
See my point on how vital it can be to convert to those unit files as
soon as possible and that's just one application stack built on top of
five components.
The main problem here are the maintainers that do not respond to the
bug report or when I ping them on irc because then I dont get the
status from them which in turn FESCO wont either since I deliver them
progress report on each meeting along with keeping [1] updated.
I'm not asking for much of maintainers *valuable* time (apparently
others people time does not matter here. We are all busy with work
life and what not, expecial those of us that are not working full time
on fedora and are doing this work on our freetime and apparently it's
not considered rude on maintainers behalf to not respond to bug
reports to the people reporting them ) only that they respond if they
are looking at this or not or will be after a week for that matter.
( which is like minute of two of their precious time ).
When I ping people on irc, private emails bugs and what not it's not
because I seek some enjoyment in nagging/bothering I do it because
it's necessary and the best way to get rid of me if I'm such a bother
that's to convert the legacy sysv init script because then you go of
my radar and I leave you be...
Patience padawan, patience.
Simo.
--
Simo Sorce * Red Hat, Inc * New York