Re: Q: Do you believe in Supercow?
On 15360 March 1977, David Kalnischkies wrote:
Old codebases usually do not attract many new people.
Well, yes, but what is that supposed to mean?
Not much more than something you probably already knew.
APT had at least one (serious) sort of rewrite (cupt) which isn't
exactly overrun either desperate it being many years younger.
I do remember cupt, yeah. Never really used it.
Debian itself is pretty old and we are talking constantly about how to
attract new blood beside keeping the current happy. Taking your response
at face value means that this is pointless. Makes me wonder why we talk
about BTS, git-debian-archive or whatever through.
Debian is not one codebase. A big collection of many different ones.
And something not being easy is no reason to keep trying. Not many new
people still means some.
And if we despair because it is hard, well, that won't make it better.
"Renewing" our toolstack and making it easy (easier) to contribute in
the small and the big ought to get more pople involved. And may overcome
the "its old, I want something new and fresh, I get more satisfaction
out of that" stigma. And some of those new ones are bound to dig into
the "hard" cases.
Also, ok, to give you another reason why not many people may step up
to do more on things like apt or dak:
They are both kind of high profile and in the middle of everything.
Break em and you break half the world.
That is scary. That gets you lots of attention. Not always from the most
friendly sort.
> Or better yet, an idea on how to change that?
if you find out how, tell me what you did, so I can repeat it for dak.
What has seemed to have worked in the past is that the entire team left
and out of the dark emerged a new team to salvage the pieces. Seems to
have happened a few times in apt. I have seen it happening in
aptitude.
Not sure if that is to be emulated and from the "outside" it seems like
a waste of resources – beside being a high stacks gamble.
Yah, thats not something to recommend.
As you are asking about dak specifically, based on my experience in the
past (but that is some years ago) it was way too hard to get a local
test instance up and running. I think/hope that changed, I would like to
revisit the {Contents,Packages,…}-all thing eventually.
Well, I am sure we aren't the easiest to setup, but hey, we do have some
gitlab ci run on our dak, and that stuff includes setting things up all
the time. And our setup docs also got much better.
> 2. There are glimmers of dissatisfaction hidden between "bikesheds",
> "curl|sudo bash" and mentions of heretic tools like npm/yarn/cargo/….
> Given a timemachine, infinite funds and unquestioned management powers,
> what would you have made APT developers do one/two/five/ten/twenty years
> ago to make you happy now?
Whats it this year with people handing out lots of magic? Is that a new
trend I missed?
A certain DPL candidate has in his platform: "We should look why people
chose something different than Debian. And see if we can enhance Debian
to provide the features, while balancing it with our current users."
Can't think of who that git has been. Still, that is based in reality,
not in magic with infinite funds, suddenly all developers agreeing to
whatever i want and even doing the work for me. I mean, that's sure a
dream, but I am not that much gone to expect such a thing :)
I am not deep enough in the apt mud to tell you what would need to have
gone different whatever time ago to make us more happy now.
Most people aren't experts in climate change but still have an idea what
could be done (if it will work or is sensible is a technical detail).
I have an idea what can be done in climate change because I get told all
the time what needs to be done if I happen to look outside Debian lists.
I am not getting told a lot of what can be done in apt all the time.
"curl|sudo bash" is even a quote from your platform, so you surely have
some opinion, broad vision or whatever. The "magic" setup is just there
to prevent you from dropping an item from the list just because it seems
too hard, out of scope, out of fear it might be silly and so on.
Well, yes. I do have opinions. But those are of the general sort and
seen from the admin user side.
That is, I like the security "theater" a package manager adds around
stuff that gets installed. In our case, with the gpg dances from the
maintainer and the archive. Which is entirely missing in the curl|sudo
bash case.
I also like the "Simple to get rid of the installed stuff"
again, somehow missing too in curl|sudo bash.
I sure also like the uniformity coming out of such a package manager.
"apt install foo" and I know what will happen, in which order, and know
where I can "jump" in if something goes wrong.
Or, to not just jump on the curl users, all that is missing in many
"managers".
So I sure know what I like. But that is a far, far different thing from
knowing what actually needs to go into such a package manager.
Sure, going from my platform and what I wrote in this list, I could go
and tell you that you should have integrated ways to replace curl|sudo
with (say) "apt install http://some.url.with.a.script"; and nodejs
whatever with apt foobar, and it would make me happy.
Actually, no, it wouldn't. As that would mitigate just one of the above
points. And miss the rest.
> 4. There is always the lingering question if Debian might become more or
> less important in the future, but asking you that seems unfair as most
> of you will not have a crystal ball. So my more realistic and totally
> technically objective question is: Do you believe APT will be more or
> less important (within Debian) in the future?
I've not seen a real replacement proposed, so it will at least stay what
it is. If it gets more or less depends on what you apt people do with
it.
While objectively true, it is a very boring response even if consistent
with your platform in that you address the Debian version of that
question in it already.
The "apt people" part is annoying me through as it suggests the apt
people are an isolated pocket. In other threads there is talk about
the close maintainer-package relationship, if/how that should be
broken and/or even sort of abolished by granting all DDs free reign
over all packages in a central version control system. Yet, the things
which are truely Debian as even their version number indicate that
they are only useful in Debian are not (yet) the responsibility of all
of Debian but of some people only.
That wasn't my intention, sorry. Thats, .. habit, I guess. It is our
current view of things, and it will take a while to change. Even if I
can get started on it (and assuming it won't run into a big wall but
proceed nicely and speedily), it will take much longer than one DPL term
to get to the goal.
--
bye, Joerg
Reply to: