[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Discussion - non-free software removal

On Fri, Nov 15, 2002 at 05:47:39PM -0500, Branden Robinson wrote:
> On Fri, Nov 15, 2002 at 03:50:38PM +1000, Anthony Towns wrote:
> > On Thu, Nov 14, 2002 at 12:07:23AM -0500, Branden Robinson wrote:
> > > On Thu, Nov 14, 2002 at 12:15:32PM +1000, Anthony Towns wrote:
> > > > On Wed, Nov 13, 2002 at 06:09:55PM -0500, Branden Robinson wrote:
> > > > > Non-free certainly costs *something*, even just at a concrete,
> > > > > shoveling-bits-around level.  What *is* that cost?
> > > > Given contrib, nothing.
> > > That has to be incorrect; it is impossible that the marginal cost of
> > > non-free is zero, though it certainly may be very small.
> > Well, you have to type "non-free" in a few places. That'd make it less
> > than the cost of just about any given mail to a mailing list.
> Hmm, but the same would be true of a "non-free enabler" package that we
> might ship in contrib and which would add all sorts of non-free package
> repos to one's sources.list.

I have no idea what you're talking about. The topic is the cost of
maintaining non-free within Debian, not getting non-free software on
your Debian box, nor maintaining non-free outside Debian.

mentions of non-free in katie.conf:

]  95:          non-free;
] 166:          non-free;
] 201:          non-free;
] 251:          non-free;
] 282:          non-free;
] 328:          non-free;
] 376:          non-free;
] 525:  non-free
] 526:  {
] 527:        Description "Software that fails to meet the DFSG";
] 528:        MeetsDFSG "false";
] 529:  };

Now, obviously, there's more to it than that, but not a whole lot
more. It's trivial enough to get completely lost in the noise.

> > > I'm not talking about human, intellectual labor.  I'm talking about
> > > network traffic and disk sectors.
> > You should be: human, intellectual labour is the most costly and limited of
> > our resources.
> It has the disadvantage of being fiendishly difficult to quantify.

That it's difficult doesn't mean it can be ignored, or that it's not worth
doing anyway.

> Of
> course it's the most important factor, but it seems to me foolish to
> *not* grab what easily quantifiable data we can,

If it's the most important factor, then it would be foolish to get
distracted by less important factors just because they're easier to deal
with. See also http://subversion.tigris.org/project_faq.html#bikeshed

> > > You're the second person today (the other was Josip Rodin) whose done
> > > statistical analysis that refused to consider non-free separately from
> > > contrib.  Why?  
> > Why not? It makes the costs look larger, which is the appropriate direction
> > for the error to go in, no?
> That depends on whether one is focussed on disk and network consumption
> in a problem in and of itself, or whether one is concerned about the
> size of the task of replacing infrastructure elsewhere.

If "one" is concerned about the latter issue, "one" should have asked about
it. It's entirely separate.

Yes, the disk space issue would be about the same, and yes, the bandwidth
issue would be about the same. What wouldn't be the same is the effort
required to setup and maintain the archive -- the human effort required,
which I think we just agreed was the most important factor.

_That_ requires:

	(a) finding a well connected machine that can be used
	(b) setting up a new dak repository (dak are the scripts used to
	    manage the archive for those playing along at home, aka katie)
	(c) maintaining the new katie repository and keeping it in sync with
	    dak CVS etc
	(d) probably setting up a new debbugs or bugzilla system, keeping it
	    in sync with CVS and removing the Debian-centricity yourself

None of these are trivial. The software related ones are probably
downright difficult if you're not experienced with them, and the main
people who are experienced with them are the current maintainers --
and having them work on it would be an obvious distraction from the work
going into Debian itself.

> > To put it another way, in terms of disk space and bandwidth the cost
> > of non-free is the same as that of two days' updates to unstable. The
> > combined cost of contrib *and* non-free in those terms is about three
> > days' updates to unstable. (On average. On big days, a single day's
> > updates uses more space and bandwidth than non-free in its entirety)
> Neither I, nor John Goerzen as far as I can tell, has forwarded reduced
> resource consumption by the Debian mirror network (including the master
> server) as a major compelling factor.

> > > > > Non-free certainly costs *something*, even just at a concrete,
> > > > > shoveling-bits-around level.  What *is* that cost?

You've certainly been focussing on it. If not because it's a reason to
support the GR, why? As a distraction to confuse the issue?

> > > > non-free software pays its own way by putting a smile on Herbert's face
> > > > when he plays whichever games he listed,
> > > Do the frowns it presumably puts on other developers' face count for
> > > nothing?
> > No one's required to expend effort on it, except to answer a single
> > question on installs, so no, I don't really think so.
> What's effort got to do with it?  

> > > > > Non-free certainly costs *something*, even just at a concrete,
> > > > > shoveling-bits-around level.  What *is* that cost?

> What about people who feel that we are
> diverging from the goals set forth in our Social Contract by packaging
> anything we can get our hands on, irrespective of its license?

Perhaps they should have read the social contract more carefully, since it
explicitly discusses that.

> > > > Considering how little it costs, it doesn't take much to pay its way.
> > > You seem to be disregarding the opportunity costs of keeping it around.
> > I'm yet to see any direct evidence of them -- even anecdotal -- and my
> > personal experience leads me to think that there aren't any, certainly
> > none that aren't outweighed by the benefits of having non-free readily
> > accessible.
> It is your contention that no one has been been motivated to make a
> piece of free software better because a competing piece of software was
> non-free?

That doesn't tie the question into Debian's distribution of non-free,
which means it's obviously not my contention.

There are plenty of bits of software that have been rewritten at least
in part because the competitor is non-free; mutt replacing pine, postfix
replacing qmail, gtk replacing qt, openssh replacing ssh, etc. Funnily
enough, that Debian distributed this non-free software anyway doesn't seem
to have been enough of a disincentive to stop any of these from happening.


Anthony Towns <aj@humbug.org.au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.

 ``If you don't do it now, you'll be one year older when you do.''

Reply to: