[ A large portion of the blame for the blow-up here is definitely mine
- see further down for more details (long)... ]
On Tue, Nov 10, 2015 at 10:17:12AM +0000, Neil Williams wrote:
>On Mon, 09 Nov 2015 18:40:29 -0800
>Russ Allbery <rra@xxxxxxxxxx> wrote:
>> This is a very difficult situation because many of the reasons why
>> Daniel's relationship with the project have not been good are
>> long-standing issues that the project addressed in various ways, in
>> some cases years ago. The relationship here has been troubled for
>> quite some time, but the various people involved have (I think with
>> mostly thoughtful restraint) have not repeatedly raised them.
>> Therefore, a lot of people who saw that thread have no idea the full
>> history of what's going on here and say things like that Daniel was
>> attacked for no reason. Which, whatever one might think about the
>> past issues, is not correct.
>> I'm very proud of a lot of people for not jumping into that thread and
>> defending themselves and arguing with the characterizations presented
>> there. This is quite a significant improvement over the way I think
>> this would have played out even five years ago, and while it's very
>> hard to stand aside and see angry things being said without full
>> context, I think that's been wise.
>It's led to a rather one-sided set of public reactions - from those who
>are unaware of the context and/or in a panic about the consequences for
>them of Daniel's reaction.
>> The main question here is how much we want to raise a bunch of old
>> tensions in order to make a public statement in response here. I'm a
>> bit dubious that it's going to do much good, and whether an extended
>> public discussion of who is right and who is wrong will actually
>> accomplish anything.
>Some feedback / positive input in public and from someone of
>long-standing in the project to say that a lot of the angst is lacking
>this context could be helpful. It's not easy to write a response which
>would do that without causing immediate demands for full disclosure and
>another round of picking over the bones of past disagreements. Some
>time may need to elapse.
>What alarmed me more was that Daniel (apparently) has sufficient
>control over the infrastructure of the live support that he can
>threaten to completely disable it in a timeframe of his choosing. It is
>this which has caused the largest reaction from users of debian-live.
>The actual arrival of code to replace live-build was the trigger. Some
>level of reaction from Daniel was expected but there was no expectation
>that the debian-live project itself was at risk. Indeed, we expressly
>used live-boot and live-config in the wrapper code because those
>By having this situation where someone at odds with the project was
>left in (sole?) control of the infrastructure of debian-live - probably
>because his access to Debian infrastructure was curtailed - it left the
>users of debian-live exposed.
>It may be worth considering that before the project accedes to making
>the output of a team a part of the "official" release, the
>infrastructure of that team needs to be under the control of teams
>within Debian. More so when the leader of such a team already has a
>fragile relationship with the project as a whole. Yes, accepting the
>work in an official capacity can be a large step towards repairing the
>relationship, it can also put users at a disproportionate risk and put
>developers in a position where it is all but impossible to make
>necessary changes. In effect, the ability to make such contributions
>official presumes the existence of a mediator and gatekeeper who
>ensures that the project and the users are equally protected. It is
>this person (DD) who would then have access to the Debian
>infrastructure on behalf of the team and be responsible for the
>processes which have been so contentious with Daniel - coordination of
>dates for releases, prevention of last minute changes to the builds,
>urgent but coordinated bug fixes and the timely addition of new support
>as required by other parts of the project.
>The vmdebootstrap support for live images is, correctly, experimental
>at the moment but live-build remains in Debian and can continue to be
>used whilst the live-wrapper images go through more testing. (Yay, we
>actually have a test suite for live images this way too.)
>> My personal feeling is to thank Daniel for all of his hard work over
>> the years for the probject and wish him luck with all of his future
>> endeavors and just leave it at that. We'll take a fair bit of bad
>> publicity for a while from people not aware of the history, and then
>> it will blow over.
>I'm happy with Iain's choice of new name for the support but I should
>have thought about it more when it was considered (briefly) within the
>sprint at the minidebconf. I'm very grateful for Iain's contribution
>and I fully support his work on replacing the code of live-build in
>the creation of official live images for Debian.
The suggestion of the "live-build-ng" name was mine, in an
off-the-cuff, tongue-in-cheek moment late-on during the sprint on
Friday. With hindsight, I should have made it more clear *immediately*
that it would be a bad name to use, for exactly the reasons that we've
seen come out here. But I can't blame Iain for following through and
using it. Iain and Neil achieved a lot in a short time, and were very
happy to show off their progress.
As a bit of background: *I* have been personally hoping (and pushing)
for a functional replacement for live-build and friends for some time,
for generating official Debian live images. I'll try to explain this
as neutrally as I can, but apologies if my frustrations intrude too
obviously - this has been festering for a long time.
Daniel and a small team of other contributors have clearly spent a lot
of time on the debian-live tools, but IMHO those tools are bad in a
number of ways. Live-build in particular is *very* fragile, and
difficult to configure. As the person who has been using it to make
official Debian images for some time, it's been a hateful
experience. I've been seeing random breakages on a regular basis, with
very little in the way of useful debugging output. Simple rebuilds for
each stable point release with no config changes barring version
numbers have often failed, with nothing logged - typically just a
missing .iso image at the end of a build being the only clue as to
failure. On an already-busy release weekend, the last thing I wanted
to be doing was debugging build failures. The internal design of
live-build makes it incredibly flexible, but also incredibly
brittle. I'm not going to make this mail even longer by going into
gory details about that here - ask me more separately if you like.
We've had a failing relationship between Daniel and the rest of the
project for a while, as others will no doubt be aware. For a very long
time, debian-live has been a personal project controlled by him. Over
the years, I and others have approached him about the options for
making it a more central official project, including things like
building official images on official project machines. For various
reasons, those discussions failed and "official" images were produced
on his own machines. After several releases with no noticeable support
from the live team (live images appearing days/weeks/months late), I
started work on using debian-live on our build machine pettersson to
make and release our official images instead of relying on Daniel to
work with us on release weekends. (We'd been trying to include him in
release planning for ages, but that failed too.) Modulo struggles with
it, I've been building images for Debian using debian-live software
for the last few years.
I've been making small fixes and tweaks to the software I've been
using over time, trying to make it simply work acceptably and
reliably. When Daniel found out I was doing this earlier this year, he
aggressively threatened to stop supporting us altogether. On top of
that, we've been missing important features for a long time (e.g. UEFI
support) with no sign of those ever coming. The minor changes needed
to make working Jessie images never made it into the archive before
release, and I could see no plan to do anything about that. Things
have been deteriorating for a while here. Ben Armstrong has been
acting as a liaison with Daniel for a while now, and I thank him for
his sterling efforts to try and make things work.
I started discussing these issues with friends a while ago. Back at
DC15, Neil Williams and I spent a very productive few days working
together (me describing fairly poorly, him implementing) on adding
more support into the existing vmdebootstrap tool that would allow it
to meet our needs. The recent vmdebootstrap sprint last week involved
more plumbing and auxiliary scripts that very quickly added the extra
logic needed to make real, usable live images. That's where the new
packages have come from.
I will say that I've been remiss in not talking about this more
publically before - instead I've been concentrating on code and hoping
that things may eventually resolve themselves without arguments.
Instead, things got worse. Sorry for that, this was bad judgement on
my part and I've been shying away from this coming fallout for too
In terms of the future, I don't honestly see a good way back for
debian-live as an *official* thing in Debian. We *do* need good live
images, and I'm happy to see that we have people working *within
Debian* to make that happen now. Neil and Iain have taken some flak
for the work that they've been doing to help here. I'm happy that
they've put in a lot of effort here and we now may have a viable
future path for functional live images. Iain never really signed up
for all this - he joined in as he wanted some working live images for
the Hamradio blend. ISTR other people have joined in on core Debian
work in similar circumstances over the years...!
>I was not expecting Daniel to react in a way that seems expressly
>designed to cause maximum panic in the users of debian-live nor for him
>to have the (assumed?) ability to threaten the future of the
>debian-live project. I was also expecting that someone within
>debian-live would either be able to speak up or take over or attempt to
>mediate, despite Daniel's reaction.
>It also didn't help that I've been (& still am) off sick since shortly
>after my own talk at the minidebconf and Steve has been catching
>up on sleep after organising the minidebconf, leaving a rather thin
>response from the debian-cd team.
Steve McIntyre, Cambridge, UK. steve@xxxxxxxxxx
Can't keep my eyes from the circling sky,
Tongue-tied & twisted, Just an earth-bound misfit, I...
Description: Digital signature