Re: [gentoo-dev] Suggestion for User Feedback

[Dan Armak]

On Monday 30 July 2001 22:17, you wrote:
> On Mon, Jul 30, 2001 at 07:40:19PM +0300, Dan Armak wrote:
> > Just an idea: what if users (those who agree, of course), when doing
> > emerge rsync, transmit to the gentoo server anonymous info on which
> > packages they have installed? If not, maybe at least the developers?
> >
> > The rationale: I have twice now faced a problem where a complex ebuild I
> > built and commited quite some time ago wasn't functioning. And I said to
> > myself, this is a very important program who's ebuild isn't working! If
> > this was a generic problem I would be flooded with emails from the other
> > people who use it! And I wanted to find these other people who used it,
> > but there were none among the gentoo developers, and I couldn't contact
> > any of the users.
>
> This is confusing.  What I read (and probably not what you ment to say) is
> that when ppl try to use an ebuild that doesn't work, they will not tell us
> about it?
They will, but not through the system I propose.
We will at some point have a standard bug report feature/script. I proposed a 
complementary system which tells us when an ebuild *does* work - which can 
also be very important

>
> > This way, we'll know what ebuilds are *known* to work properly - at least
> > under some h/w configurations - and most importantly, in what
> > combinations and revisions. And it will give the users a painless way to
> > help us.
>
> Many people find it quite painful to share details about their systems.
> They tend to exagerate it into all kinds of personal invasion.  Even when
> there are perfectly good arguments that it is not.  In something like this,
> what people think it is has more weight than what it really is.
Well, we'll juts turn it off by default, and have the user read an 
explanation and decide whether to turn it on.

>
> > Finally, we'll also know what packages are most often used, in what
> > combinations, and thus (as far as we work for our users) what areas to
> > concentrate upon.
> >
> > What do you think?
>
> Other than my automatic rejection to collecting info, Someone wil have to
> spend a good deal of time building the collection system.  And right now I
> think energy could be better spent elsewhere.
The info would be fully anonymous, as is the rsync protocol. I'd want the 
info that e.g. such and such combination of ebuilds works together in such 
and such revisions. Like getting the directory tree of /var/db/pkg without 
the files. The used make flags, etc. could also be useful. The user could 
customize what info to report.

As for building the system, the user-side thing would be a couple of pages 
long at most, since we already have an rsync implementation in emerge.

The server-side thing could be more or less complex, depending on how 
powerful the system would be. Making a database with dynamic online 
frontends, queries, comparisons etc. etc. would certainly be quite difficult 
but we don't necessarily need all that functionality.

>
> I can see how the information would be useful.  And I would suggest a
> script that users can run, view the collected infomation, and upload
> somewhere over anything automatic.  By putting the details into the hands
> of the users, and letting them see exactly what they are going to send
> (they wouldn't have to look at the contents, but we should give them the
> chance) before they send it, people will likey be more comfortable about
> sharing the info.
As you say.

>
>
> I imagin I just really confused things....

hallski posted much the same questions, see my reply to him.

-- 

Dan Armak
Gentoo Linux Developer, Desktop Team
Matan, Israel