Sven Neumann wrote:
> Your mail really confused me as I don't seem to be able to follow most
> of the points you brought up. Dave, is that really you who wrote this??

First off, let me assure you that it was me :)

Right up front I said:

> David Neary <[EMAIL PROTECTED]> writes:
> > The problems as I see them are...

That was intended to emphasise that this is my take on things,
and is, as with everything I say, open to discussion and quite
possibly wrong.

I will try to clear up some of your confusion for you :)

> > 1) User feedback on the development series is poor
> > 2) Documentation is poor
> > 3) Our release cycle is poor 
> > 4) UI is not a priority.
> I am very surprised about point (4). We have put a lot of effort into
> improving the user interface. I would even go so far to say that UI
> has the highest priority.

From my perspective, I've seen discussions in recent months on
modifying shortcuts to conform with the HIG, reduce the
complexity of the install process, setting the menu on the image
windows on by default, and on and on. There have been 4 or 5 of
these arguments and in every case your position has been to stick
with the status quo. People have said to you "users find this too
complex. I find this too complex." and your answer has been "Some
people find it useful".

I hope that you don't think I'm picking on you here - Raphael has
also tended to favour the status quo in usability bugs on
bugzilla. One question that springs to mind from what you say 
above is who is "we"? I've not been involved in many UI
discussions, although I have made a first contact with
[EMAIL PROTECTED], and hope to have a mail sometime this week
requesting screenshots for usability evaluation...

> Also I have to disagree with point (1).
> There is substantial user feedback on the development series. You
> probably don't read the gimp-user mailing list and don't follow the
> discussion on various public forums whenever a release is done. Other
> developers seem to do that and all this feedback is going back into
> the development. I really don't see your point.

Up until quite recently (that is, during the first 2 years of 1.3
development) anyone who wanted to open a bugzilla report against
1.3 was told they couldn't - that the series was unstable, full
stop. So up until quite recently, there wasn't any way for users
to give feedback in terms of bug reports. Also as you may have
noticed, the devel list has gotten very quiet recently (well,
until you proposed that the next version be called 2.0 - that
might actually have been a good move :). 

As you say, I am not subscribed to the user list. Perhaps I
should be. I suspect I'm not alone in not subscribing. But the 
idea of the separation of lists was that I didn't have to be. 
Perhaps that day is gone, and I should just subscribe to the 
user list and be done with it. In this case, if all the 
developers are subscribed to the user list, why would we need 
a devel list?

> > 1) Not enough users use bugzilla to report bugs
> Bugzilla is the only way to report bugs. Whooever reports bugs uses
> Bugzilla for it. A few people need to be shown the way to it but in
> the end everyone uses Bugzilla.

Bugzilla is a quite complicated tool. I love it. As you say,
anyone who reports bugs uses bugzilla. That means we miss lots of

> > 2) Not enough developers use Bugzilla to find out what bugs need
> > fixing
> > 3) Not enough developers hear user complaints
> I believe that all active developers use Bugzilla. Mitch and me spend
> a substantial amount of time with Bugzilla and I have the impression
> that other developers do the same.

"Use bugzilla" != "add occasional comments to bugs and maybe fix
a few". As I'm sure you've noticed, 90% of comments on open gimp
bugs are from either yourself or Raphael.

A quick trawl of bugzilla shows that there are 182 bugs with
comments from Mitch in the last 6 months, compared to 691 from
yourself, 479 from Raphael, and in the region of 80 for me
(because I changed emails in the last 6 months of my bugzilla
account it's a bit hard to figure out).

This just to show that like in terms of code contributed, where
there's a huge gap between yourself and mitch vs the rest of us,
in bugzilla it's yourself & Raphael vs the rest of us (I suspect
this is because we use bugzilla wrong).

> > - Mail alias - [EMAIL PROTECTED]
> This one is dead, noone is using it. It's only used for syndicating
> Bugzilla changes.

Not no-one is using it. I use it to do a quick filter on bugs I
might find interesting. I delete 99% of the mails as they arrive. 
I actually use this more than bugzilla (where "using bugzilla"
means getting ownership of bugs, and fixing them or passing on 
ownership to someone else).

> > My first proposal would be to do a reverse split of the users
> > and developers mailing lists - get everyone talking to one
> > another. It will certainly annoy people because of increased
> > traffic, but I think it'll be worth it. We have to face up to the
> > fact that after 3 years without a major release, and only 14
> > active developers, the GIMP is a small project. Step 1 is to get
> > people talking to each other.
> This might be a good idea but I am afraid that it will have a negative
> effect. I fear that users will stop subscribing to the list because
> there is too much technological discussion and I fear that developers
> will be annoyed by too much ever-repeating newbie questions.

Perhaps the annoyance will manifest itself in the new FAQ that
people (including myself) intended writing a year ago? And the
devel list's been so quiet recently that the tech discussions
probably wouldn't be noticed. But if they are, maybe we'll start
getting contributions from people who want to help again...

> > Proposal 2 is to either do away with [EMAIL PROTECTED] or de-spam it.
> We need [EMAIL PROTECTED] Despamming it is of course a good idea and I
> wonder why Yosh has disabled the spam-filter again. Getting rid of
> [EMAIL PROTECTED] as a syndication of changes to Bugzilla is definitely a
> very bad idea. It would cause unseen changes to bugs and new
> bug-reports being ignored.

Yet again, would you mind saying who is "we"? And if we used
bugzilla properly, there would be no missed bugs... I hate to
insist on this, clearly I'll need to explain it a bit more (will
do, below).

> Have a look at the stats, there are no new bugs floating around.
> Usually bug-reports get answered in less than 24 hours. We trimmed
> down the number of open bugs to a half during the last year. If there
> is some success story in GIMP development, then it's Bugzilla. The
> main cause for this is the fact that developers get mail from Bugzilla
> whenever a new bug is reported or a report is changed.

Right ideal opportunity to explain how bugzilla *should* work.
Let's look at the currently open gimp bugs - there are, in total,
391 bugs. Which is, as you point out, doiwn from over 600 at one
point. I even had a little bit to do with that, I think... 

Now, let's look at the "Owner" field - 1 bug is owned by
grosgood, one by jimmac, 2 by mitch, 2 by Raphael, 2 by rockwlrs,
and 1 each for sven and yosh. That's 10. Leaving 381 bugs owned
by that well-known gimp developer [EMAIL PROTECTED]

The lifecycle of a bug *should* be that a bug gets reported
against a module, which is owned by the module owner, who knows
who works on what parts of that module (for example, a bug
against the plug-ins module would land in the mailbox of someone
who knew that adam owns the psd plug-in, but that myself or
Maurits might be interested). This person doesn't need to be
technical, they need to be good enough to screen out the
NEEDINFOs, the NOTABUGs, the INVALIDs. Otherwise their job is to
turn unconfirmeds into news, and to pass the bug on to the person
(or group of people) interested. 

At this point, the person to whom the bug has been assigned
either accepts the bug (effectively agreeing to fix it), or
re-assigns the bug to someone else. When the bug is fixed, it is
reassigned to the original reporter for verification and
eventually closure. 

That is what is supposed to happen in bugzilla. What actually
happens is kind of similar, in a haphasard way. Usually a bug in
one of the file format bugs gets adam's email stuck onto it by
someone, and he ends up having a look. Or I notice a keyword in a
mail I get at [EMAIL PROTECTED] that might be related to a bugfiw I
botched up. And so on.

> > I would prefer to set up a system of module owners, who take
> > possession of bugs, and send them on to the people best suited to
> > fixing them. These people wouldn't even need to be technical, they'd
> > just be required to do a first-level filter (invalid or notabug or
> > needinfo bugs) and send the real bugs to the people most likely to
> > fix them.
> Of course a good idea in theory but facing the fact that there are
> only very few active developers, it doesn't seem to be doable. If you
> can hire a team of 5 QA persons for this job, I'd vote for your
> proposal.

I'm glad you agree there's a problem with the number of people
active in gimp development. As I've said a couple of times, the
module owners don't need to know how to fix the bugs. This seems
to me like a great way to get people involved in the gimp. I know
that fixing bugs was how I got involved... I just started
browsing bugzilla, picked one that looked fairly easy, and
attacked. So maybe there are 5 people who aren't too confident
diving into the core, and would like to feel their way around
with bug reports?

> > Proposal 3 is to try to persuade the Win32 guys to come back to
> > the main gimp mailing list. 1.3 should be buildable out of CVS.
> This has happened, the win32 development is discontinued and
> subscribers have been asked to join gimp-developer and/or gtk+-devel
> instead.

And yet I can't remember the last time I saw a mail from Hans or
Tor. And I know that they had gathered a group of Win32 people,
none of whom seem to have appeared on this list. And I have yet
to see a screenshot of the GIMP 1.3 running on Win32 (except via
an XServer, which I've done :).

> > Proposal 6 - allow people to submit bug reports without a
> > bugzilla account. I would like it if Bugzilla could get their
> > email address from the first mail they send to the portal,
> > sign them up and send them a password, but it doesn't. As a
> > technical problem, is this possible? Or could we have a mail
> > alias to which mails (which pass a spam filter) get converted
> > into bugzilla reports, with the e-mail information in the body of
> > the bug report?
> This would mean more bug-reports without a working email address of
> the bug reporter. IMO a very bad idea since experience shows that
> these bug reports cause a lot of work and often need to be closed as
> INCOMPLETE or INVALID in the end.

What's wrong with using the From: or the Reply-to: addresses?


> > be "gimp crashes, go to gimp web page, nothing about bugs on
> > first page, scroll down 4 pages, follow "Submit a bug!" link,

> There is info about bugs on the first page of www.gimp.org. I don't
> know what you are talking about.

Screenshot: http://bolsh.dyndns.org/~dave/Screenshot.png

Where do you see any mention of bugs, or support, or anything a
user might be tempted to follow in the event of a problem?

I thought the mention of scrolling above made it clear what I

> > OK - so that's it. Food for thought. Basically, since 1.2 the
> > size of the gimp community has been shrinking. We don't have any
> > documenters, we don't have many testers, we don't have many bug
> > fixers, we have very few web developers, we have a couple of
> > artists, and we have maybe a dozen active developers. Something
> > needs to be done to change that, or the gimp will never see a
> > major release with gegl (whatever version number it will have).
> Your mail really confused me as I don't seem to be able to follow most
> of the points you brought up. Dave, is that really you who wrote this??

It doesn't surprise me much that your take on this is different
to mine. I hope you don't take offense at this, but while you're
a great developer, no-one could ever accuse you of being a good
politician :)

Part of the idea behind this is to establish some kind of
structure to how we use the tool available to us in a manner
which makes it easier for people to contribute to the gimp. As
you already contribute immensely to the gimp, I wouldn't expect
you to see any problems with the way things are. But when I spend
more of my free time building the gimp or installing dependencies
than I actually spend coding, I consider that a problem.

Part of the problem is communication. In the past 3 years, there
have been numerous surprises after doing a cvs up - sometimes
it's a version jump in automake, there was the version jump in
gtk+, then there was fontconfig, there was the disappearing
toolbox way back when, then the reintroduction of the mmx stuff,
and so on - at various moments during the development of 1.3,
people's build environments have been broken overnight with no
mail to say it was going to happen, or why. I remember a big
argument about whether the 1.3 version of GTK+ was really
necessary - in my opinion that jump was badly managed, and
probably resulted in a half dozen people stopping building cvs

But it also goes the other way - I remember arguments over
plug-in tools and mmx in particular, but there have been a few
occasions where people have submitted stuff to the gimp which has
subsequently been backed out. This is not to say that the
decision to back stuff out was necessarily bad, but if I remember
them it's because they were badly handled.

I'm sure that you've personally been frustrated by the lack of
help that's been around for yourself & mitch. But the reason
there's been a lack of help is because the people who are likely
to help are so often opposed when they propose stuff.

My idea is to discuss a change in how the gimp project runs
itself. The changes would include spreading around the
responsibility for things like bugzilla in a more official way -
someone would be In Charge of the plug-in bugs, someone else
would be In Charge of the UI bugs, and so on.

To me the benefits of these types of changes are self-evident -
they partition up things to be done into bite-sized tasks,
allowing people who have a couple of hours a week free (like me)
to be of benefit to the project in a more targetted way. They give
us a chance to start rebuilding the kind of community we had when
I started working on the gimp. They take a bunch of work away from
yourself and mitch, and spread it around.

They also allow people who want to contribute to the gimp to do
so more easily. Getting the gimp developpers and the users
talking with the intention of having a better gimp, and a better 
community, is a good thing. We need to spread responsibility
around, and that includes responsibility to decide what does &
does not go into CVS. 

Hopefully this will develop a bit more and people's positions can
be cleared up before camp, I would like to spend some time at the
camp discussing these types of issues (communication, and
distribution of responsibility). Of course, if we can all agree
before camp, great. More time for beer.


       David Neary,
       Lyon, France

Attachment: pgp00000.pgp
Description: PGP signature

Reply via email to