On Tuesday 19 January 2010 12:39:39 Murray Cumming wrote:
> On Tue, 2010-01-19 at 08:43 -0300, Jeff Moe wrote:
> > 
> > I don't have to be here for years to see how things are managed and
> > where the problems are. 
> 
> Things are clearly in progress. It's apparently a move in the right
> direction,

I'm not quite sure they are. Didn't the server just get moved *to* an ISP that 
can't fix NFS mounts on weekends? What ISP is this?

> even if it's not everything you want. How about letting it
> settle down just a little bit and then calmly suggesting how to avoid
> problems in future.

People have been preaching patience (not just to me, to everyone) since I 
landed here and before. Why should we be patient? Why can't we demand things 
work like they do everywhere else? I'm sure google/apple would love for us to 
be patient for the next 3 years.

> I am sure that Tero and his colleagues would actually like ideas that
> reduce the awkward infrastructure work.

I've already suggested *many* ways for things to be improved:

* Copy Fedora Infrastructure Standard Operating Procedures, in which I track 
down a Fedora admin to get info, explain how they do things, and give links for 
more info:
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023329.html
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023337.html

* I also sent a list about the same issue to council, providing more info about 
how other distros handle such things. (offlist, but the council sent it to the 
list before checking with me).

* I suggest using mirrors well before the most recent outage. Not only do I 
suggest it, I go on a actually set up a mirror, fully document how it is done 
in the maemo.org wiki, and give full root access to a Maemo admin. Later on 
this mirror is heavily used during a major maemo outage. I also within hours 
set up two additional mirror servers in distinct data centers:
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023337.html
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023392.html
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023425.html
http://espejo.freemoe.org/
http://wiki.maemo.org/User:Jebba/Espejo
http://wiki.maemo.org/User:Jebba/Mirror

* Suggest, and write a mini-howto about documenting server changes. It appears 
the maemo admin team cannot tell when or who has made changes to system 
configurations (!).
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023560.html

* Wherein I discover that the SDK has been silently changed and suggest that 
version numbering be used:
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023776.html
This directly lead to this thread about how the SDKs should be set up, which is 
of course vitally important:
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023781.html

* Suggest moving DNS to someone other than Nokia, due to a major extended DNS 
outage (IMHO, this should *never* happen). This could be *easily* implemented 
and would greatly increase reliability of the entire infrastructure:
http://lists.maemo.org/pipermail/maemo-developers/2010-January/023805.html

* Along with all that, I have frequently documented actual outages in what is 
currently the "proper" place for this. I believe my reports were good--in any 
case I asked if more info was needed and none was asked for. I even offered SSH 
access to confirm it.
https://bugs.maemo.org/show_bug.cgi?id=5818
Note, I haven't documented every single outage I've seen--there have been far 
more. In fact, *today* many people around the world are still reporting outages 
on talk.maemo.org.

---
If I seem a little testy maybe it's because I've been told to be patient for 
way too long. "Patience" with a server outage should be measured in hours, not 
months. If maemo doesn't find a way to solve these problems, they will be 
crushed in the end. A blogger recently pointed out along the lines of 
"mindshare is the marketshare of the future". Maemo is losing mindshare and 
developers due to these *continued* problems.

I've seen frequent mentions of how distinct maemo.org is from Nokia. But the 
reality seems that nothing at Maemo can really happen without Nokia's tacit 
approval. Is there anyone that has server access that isn't paid (directly or 
indirectly) by Nokia? Can we start getting the servers admin'd by community 
admins instead of depending upon Nokia? A first step would be to document (on 
the *outside/public*) wiki the current server arrangement. Another good step 
would be to get DNS off their servers. Until that happens, the whole "maemo is 
distinct from Nokia" is just a façade.

I once sent this analogy to someone:
Nokia built a wonderful hockey stadium, great seats, boxes, huge LCD screens. 
They gave out free tickets to the the public and a free beer token to adults. 
The players got new uniforms, clean locker rooms, and top doctors. Then the 
athletes went to play, but the equipment manager gave them styrofoam hockey 
sticks. That's basically what we're working with. That equipment manager needs 
to come up with a solution NOW, because the *GAME IS ON* and Nokia is getting 
kicked--they're down 20,000 goals (applications) to 110 after the first period.

Impatiently yours,

-Jeff
_______________________________________________
maemo-developers mailing list
maemo-developers@maemo.org
https://lists.maemo.org/mailman/listinfo/maemo-developers

Reply via email to