Re: Bug#800769: pbuilder: conffiles not removed

2015-10-11 Thread Wouter Verhelst
On Sat, Oct 10, 2015 at 04:30:28PM +, Mattia Rizzolo wrote:
> Hi fellows debian-devel@ lurkers!
> 
> On Sat, Oct 03, 2015 at 09:03:46PM +, Mattia Rizzolo wrote:
> > On Sat, Oct 03, 2015 at 02:33:09PM +0200, Paul Wise wrote:
> > > The recent upgrade did not deal with obsolete conffiles properly.
> > > Please use the dpkg-maintscript-helper support provided by dh_installdeb
> > > to remove these obsolete conffiles on upgrade.
> > > 
> > > https://www.debian.org/doc/debian-policy/ch-files.html#s-config-files
> > > http://manpages.debian.org/man/1/dh_installdeb
> > > 
> > > $ pkg=pbuilder ; adequate $pkg ; dpkg-query -W -f='${Conffiles}\n' $pkg | 
> > > grep obsolete
> > > pbuilder: obsolete-conffile /etc/pbuilder/pbuilder-uml.conf
> > >  /etc/pbuilder/pbuilder-uml.conf ce1832f09d721efe29b92ec153fa4410 obsolete
> > 
> > oh, gosh
> > All this just for fucking up with the arch:all buildd :\
> > 
> > This happened because the 0.217 was the first version being built by the
> > arch:all autobuilder, but the code that moved that file away was run
> > only for arch-indep builds, so it did not run there; the 0.218 was
> > uploaded exactly to fix that issue, but 0.217 was already in the wild
> > with that bogus file.
> > 
> > That file is supposed to be only on pbuilder-uml binary, so I can't just
> > use rm_conffiles to remove it in pbuilder, as that would blindly remove
> > it while pbuilder-uml might be using it.
> > 
> > How do you suggest to handle this?
> 
> Do you happen to have ides?
> What actually needs to happen is to hand a conffile over to another
> package, but neither me or pabs have a clue on how to do that.

The only way to hand a file (any file) over to another package is by way of
'Replaces:', *without* the Conflicts: and Provides:.

Since this is a single packaga version, you could put the file in pbuilder-uml
and have that 'Replaces: pbuilder (= 0.217)', that way the possible damage in
case you have other (accidental) file conflicts would be limited.

-- 
It is easy to love a country that is famous for chocolate and beer

  -- Barack Obama, speaking in Brussels, Belgium, 2014-03-26



Re: Bug#800769: pbuilder: conffiles not removed

2015-10-11 Thread Mattia Rizzolo
On Sun, Oct 11, 2015 at 11:03:09AM +0200, Wouter Verhelst wrote:
> The only way to hand a file (any file) over to another package is by way of
> 'Replaces:', *without* the Conflicts: and Provides:.
> 
> Since this is a single packaga version, you could put the file in pbuilder-uml
> and have that 'Replaces: pbuilder (= 0.217)', that way the possible damage in
> case you have other (accidental) file conflicts would be limited.

Though this way if a person install pbuilder 0.217, then removes it the
conffile would stay there, and that's the whole point of the bug.
I agree the Replaces is needed, but I don't see it as being enough to
deal with this... :(

-- 
regards,
Mattia Rizzolo

GPG Key: 66AE 2B4A FCCF 3F52 DA18  4D18 4B04 3FCD B944 4540  .''`.
more about me:  http://mapreri.org  : :'  :
Launchpad user: https://launchpad.net/~mapreri  `. `'`
Debian QA page: https://qa.debian.org/developer.php?login=mattia  `-


signature.asc
Description: PGP signature


Help needed talking to upstream browser developers about Debian SSO

2015-10-11 Thread Enrico Zini
Hello,

I need help to make our SSO use case for client certificates known in
the upstream browser development communities. Here is the story.

At DebConf we deployed a new experimental single signon system based on
ssl client certificates[1], and it works quite nicely. In particular, it
allows anyone to create services that authenticate Debian people, with
just a few web server configuration lines[2]. I am now aware of any
other single sign-on system that has this combination of simplicity,
security and usability.

There are still a few things that could be better, and I was kind of
dreaming that we could build from here[3], just as we recently did a lot
of work to ensure that lots of browsers in Debian could work with client
certificates[4].

However, there is discussion in the Chrome[5] and Mozilla[6] communities
about deprecating client certificate authentication. In those threads,
vocal people in support of Client Certificates argued mostly for WebID
and seem to have managed to completely alienate the Chrome developers.
Other use cases, including ours, were not really represented.

My understanding so far is that client certs seem to be better than
anything else (one enrollment step, and sites that want to use the auth
just add the CA cert and CRL to the web browser configuration).

I don't quite mind if  is removed, as long as there would be a
replacement that allows the existence and growth of an ecosystem with
shared identification, based on popular standards and easy to use and
deploy.

FIDO and crypto APIs are mentioned as alternatives which do not seem to
be there yet. I appreciate that the Chrome people have said that they
intend to wait for an alternative to be viable before removing keygen,
although I would want to be sure that they are aware of our use case.

I dream of a low-maintenance, established technology that allows a
community to set up a central authentication place, and then let a
decentralised network of web services grow around that.

Client Certificates allow to do that today: it is easy to have a central
place authenticating users and enrolling devices, and it is easy to
accept and trust those certificates without having to coordinate with
the issuing site.

I wish that this use case would be explicitly supported. I was happy to
see existing, established standards for it.

Unfortunately I do not have the time and energy to actively participate
in those discussions, so I'm asking for help.

To me, sso.debian.org is a side project that I'm working on to support
the sites I actually care about: nm.debian.org and
contributors.debian.org.

So far I have maintained server-side code and patched client software,
and now the next step would be to carefully negotiate with upstreams to
figure out what can be the future of sso.debian.org. I am losing
motivation, because despite all this work, the things I care about are
really not progressing.

This is a call for help: please help making sure that there are standard
ways for mere mortals to setup and maintain central authentication
services for their community.

I would appreciate being able to just focus on nm.debian.org and
contributors.debian.org knowing that I somehow have my back covered on
the sign-on side.

I see a few things that could be done:

 - participate in those discussions constructively, being a rational
   voice showing why and how we use client certificates in
   sso.debian.org and discussing with the browser communities what could
   be concrete futures for it.

   Talking with Chrome developers seems to require a Google account;
   talking with the Mozilla developers does not:

  https://www.mozilla.org/en-US/about/forums/#dev-security
  https://www.mozilla.org/en-US/about/forums/#dev-platform

 - package and publish django-spkac[7] on a popular platform like GitHub
   and https://www.djangopackages.com/, with a README somehow explaining
   the vision and few steps[8] for deploying Single Sign-On for a
   tech-savvy decentralised ecosystem like ours, and show it as an
   example of our use case.

I want sso.debian.org to grow without me. You may have time, energy,
expertise, further and better ideas: by all means go ahead with them.

Thank you in advance,


Enrico

[1] https://wiki.debian.org/DebianSingleSignOn
[2] 
https://wiki.debian.org/DebianSingleSignOn#Documentation_for_web_application_owners
[3] http://www.enricozini.org/2015/debian/if-you-know-a-browser-developer/
[4] https://wiki.debian.org/DebianSingleSignOn#Browser_support
[5] https://groups.google.com/forum/#!topic/mozilla.dev.security/mr_DoJGOoiA
[6] 
https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/pX5NbX0Xack/kmHsyMGJZAMJ
[7] http://anonscm.debian.org/cgit/debian-sso/debian-sso.git/tree/spkac
[8] 
http://anonscm.debian.org/cgit/debian-sso/debian-sso.git/tree/README-spkac.md
-- 
GPG key: 4096R/E7AD5568 2009-05-08 Enrico Zini 


signature.asc
Description: PGP signature