Re: [ANN] Please use Gerrit from now on for Patch Review

On Tue, Jun 19, 2012 at 04:46:26PM +0100, Michael Meeks wrote:

> On Tue, 2012-06-19 at 17:04 +0200, Lubos Lunak wrote:

>       My hope is that by encouraging the use of gerrit in parallel with the
> mailing list, the benefits will become sufficiently obvious over time
> that the old way of merging patches mailed to the list (while still an
> option) will seem to be the more annoying way to do things ;-)

>       So - I think we should give gerrit a break; and play with it and see
> what happens - though I agree the absence of a 10 bullet TLDR; rational
> has been a bit of a frustration; I'm sure we'll get past that.

>       Either way - we'll discuss this at the ESC on Thursday if you want to
> join in.

As I won't be able to join in this week, my views:

I have a very, very positive view of what gerrit is supposed to
do. As a person that asks for review, I'm constantly frustrated with
the process of sending a patch for review to the ML (git format-patch,
fire up email client, new email to list, attach patch, ...) and I look
forward to submitting with a simple "git push".

I dislike web interfaces, and from what I read, gerrit is good for
that, because one can do many (all?) things from the command line
(now, if our BTS could also have a command-line or email interface,
I'd be extra-plus happy).

However, our current setup *requires* an OpenID; is it an option to
make that optional (and allow people to e.g. use a "classic"
username+password for the web interface)?

My reasons for that is that signing up for an OpenID is quite an
involved process for people that don't already have one; at least for
people that think about the consequences (and don't like them); I
expect that would be at least the whole cryptogeek / cypherpunk
crowd, as well as the "privacy aware / control aware" crowd.

People like that (yes, I'm one of them) will balk at the requirement
of giving a third party (and anybody able to twist their arm... like
the surveillance agencies of governments) unlimited power to
impersonate them (to websites that use OpenID). So they'll want to run
their own OpenID end points; on the surface that's easy, but it
actually took me *days* and poring over standards to find a nice one
that will work with Gerrit, and I had to patch it myself.

 - gracie looks like it would do the job (just run it on your
   desktop), but it is bitrotten and does not work *at* *all* with
   recent python modules (see e.g. http://bugs.debian.org/src:gracie)

 - local-openid looks like a godsend, but gerrit won't interoperate
   with it; I now patched it (days and hours of efforts...), so
   hopefully it will become a good solution soon :)

 - prairie: alpha version last updated in 2008. ugh.

 - SimpleID seems to be nice too, and a good alternative to
   local-openid; just don't try (as I did) to use the same URL for
   your identity and for running SimpleID, it won't work.


My point is basically that it is too much of an investment for a
casual contributor... If we could make that easier by allowing plain
username+password (or exporting bugzilla accounts over OpenID? I guess
that would be *more* work), I feel it would lower the barrier to entry
to gerrit.

-- 
Lionel