Re: Gnus Git repository info and comitters: need updated password

[Ted Zlatanov <tzz@lifelogs.com>]

On Sun, 18 Apr 2010 18:35:26 +0200 Andreas Seltenreich <andreas+ding@gate450.dyndns.org> wrote: 

AS> Sounds very promising.  It uses plain CGI, so I don't see problems with
AS> the old apache.  According to the docs, it should also work nicely with
AS> the existing BasicAuth credentials for write access.  Being a "smart
AS> protocol", it should be more efficient as well.  I'm still shivering
AS> from the 7000 WebDAV-Requests git made during that short debug
AS> session...

Unfortunately it requires lots of Apache 2.x specific things, according
to the HOWTOs I found, so I'll just set up a 2.x server on git.gnus.org
(it's a separate IP so it won't interfere with the existing 1.x
servers on that machine).

On Sun, 18 Apr 2010 21:04:10 +0000 (UTC) Reiner Steib <reiner.steib@gmx.de> wrote: 

>> IMO it's best to set up an automatic publisher of the www.gnus.org and
>> git.gnus.org sites straight from Git (which is why I set up the www/
>> directory).  That will work regardless of who is the web site maintainer
>> and will take the publishing burden off that person.  

RS> I think the whole web site should be in the repository (like on
RS> savannah.gnu.org), I already suggested this to Lars (maybe a year ago).

With the git-http-backend we can set up auto-publishing so it can all
Just Work.  Do you want it automatic immediate, scheduled hourly/daily,
or manual by you and me?

You will have to change the Gnus release procedure since it's very
CVS-centric but I think it will be simpler and easier as well.

RS> Please try to make sure to update the web pages WRT cvs when doing the official
RS> switch. Do you intend to sync from git to cvs?

No, I think CVS should be shut off.  It's a waste of time to maintain
CVS today IMO.  If you disagree, tell me why :)

CVS is already disabled for commits so until www.gnus.org is updated we
can keep it running but without any new updates.  I think that's an
acceptable situation.

There's git-cvsserver as well.  I'm opposed to providing that but will
listen if there's a reason to do so.

RS> WRT to the switch to some other VC system: I'd prefer not to have a different
RS> one than for Emacs, but well...

Honestly I wish the Emacs maintainers the best with Bazaar, but I don't
want to run a Bazaar repository.  I hope this doesn't become an issue
with you or them.

RS> BTW, http://git.gnus.org says:
RS> | The comitter should also have a ~/.netrc file that says
RS> | machine git.gnus.org login yourlogin password yourpassword

RS> Is it really necessary to store the *unencrypted* password in .netrc?

I don't believe curl supports encrypting the netrc.  Maybe someone more
knowledgeable can tell us if that is planned for the future.  As you
know I've always pushed people to encrypt their authinfo file so this
situation is unpleasant to me as well.  These passwords are also
crypt-ed on the server, which is not a very secure algorithm, and of
course we're going against a self-signed SSL certificate.  I'd rather
have used SSH but that's just not an option until and if Lars approves
it.  So as with DNS and the SSL certificate, we'll wait for his decision.

Ted