Franklin Street Statement and Free Network Services

[evanprodromou]

Hi, all. I'm interested in getting some feedback from the list on the Franklin Street Statement:

http://autonomo.us/2008/07/franklin-street-statement/

...and on the Open Software Service Definition:

http://www.opendefinition.org/ossd

I'm currently working on an Open Software Service called Identi.ca:

http://identi.ca/

It's a microblogging site (think Twitter) built on Open Source software and Open Data. To add to the fun, we use a distributed social networking model, with a fun little open protocol called OpenMicroBlogging.

Things are going a leetle beet crazy with Open Source and Web services lately. An Open Source, Open Data site is the 8th biggest (or is it 7th now?) on Alexa. Reddit went FLOSS a while ago, and Ma.gnolia.com just announced their M2 project.

Second players in network-effects markets can change the game with Open Source. In a closed world, it's silo vs. silo, and the biggest silo wins. But in an open world, it's protocol vs. silo, or Internet vs. silo, and no silo is as big as the Internet. It's not a guaranteed win, but in a network-effect market, it's about the best chance you've got if you're not in first place.

Thoughts, ideas, feelings, emotions?

-Evan

--
Evan Prodromou <evan@...>

[davelab6]

2008/8/30 Evan Prodromou <evan@...>:
>
> It's a microblogging site (think Twitter)

I'm very glad you've finally implemented a proof-of-concept that a p2p
webapp can replicate the network effects enjoyed by centralised
webapps.

(For a while I've been suggesting that a few free software webapp
projects I've been peripherally involved in do this, and no one took
it on...)

I'm waiting for someone do to this to a photo sharing site (think flickr) ;-)

--
Regards,
Dave

[Ben Tilly]

On Fri, Aug 29, 2008 at 8:58 PM, Evan Prodromou <evan@...> wrote:
[...]
> Second players in network-effects markets can change the game with Open
> Source. In a closed world, it's silo vs. silo, and the biggest silo wins.
> But in an open world, it's protocol vs. silo, or Internet vs. silo, and no
> silo is as big as the Internet. It's not a guaranteed win, but in a
> network-effect market, it's about the best chance you've got if you're not
> in first place.

FYI for most networks, most of the time, there is reason to believe
that network effects scale more like n log(n) rather than the naive
Metcalfe's law of n*n.  See
http://www.dtc.umn.edu/~odlyzko/doc/metcalfe.pdf for some of the
evidence.

Therefore even in a closed world the bigger silo doesn't have that big
an advantage over the smaller ones.

Ben

[davelab6]

2008/8/30 Evan Prodromou <evan@...>:
>
> ...and on the Open Software Service Definition:
>
> http://www.opendefinition.org/ossd

> > The Definition
> >
> > An open software service is one:
> >
> > 1. Whose data is open as defined by the open knowledge
> > definition (http://opendefinition.org/1.0/) with the exception
> > that where the data is personal in nature the data
> > need only be made available to the user (i.e. the owner
> > of that account).
> >
> > 2. Whose source code is:
> > A. Free/Open Source Software (that is available under a
> >     license in the OSI or FSF approved list

This should be boolean AND instead of OR because the FSF and OSI lists
diverge slightly; there are some OSI licenses - Artistic License 1.0,
NASA Open Source Agreement 1.3, Reciprocal Public License - that the
FSF state are non-free.

http://www.gnu.org/philosophy/license-list.html#NonFreeSoftwareLicense
http://www.opensource.org/licenses/alphabetical

> > B. Made publicly available."

This should be "Made available to its users."

Requiring publication is highly contentious.

In the early "Emacs commune" days RMS decided it was unethical to
force people to publish their changes, which is why the GPL doesn't
(no link, I cuss printed books, I believe I read this in 'Free For
All' or 'Rebel Code'). The eCos Public License that does what RMS once
did, and the Reciprocal Public License that requires publication, is
criticised at http://www.gnu.org/philosophy/license-list.html#NonFreeSoftwareLicense

More recently, debian-legal have the desert island/dissident tests [1]
that, while not as important as the FSF FSD and OSI OSD, are still
worth thinking about and should not be disregarded lightly.

And most recently, the Affero GPLv3 says that the _users_ of network
software must be able to access it via the network - which I think is
striking the correct balance, although some d-l folks like MJ Ray
don't like it.

[1]: http://en.wikipedia.org/wiki/Debian_Free_Software_Guidelines#debian-legal_tests_for_DFSG_compliance

My suggestion:

An open software service is one:

1. Whose data is open as defined by the open knowledge definition
(http://opendefinition.org/1.0/) with the exception that where the
data is personal in nature the data need only be made available to the
user (i.e. the owner of that account).
2. Whose source code is Free/Open Source as defined by both the FSF
and OSI (that is available under a license in the OSI and the FSF
approved lists) and is available to the users of the service.

--
Regards,
Dave

"Nothing would please me more than being able to hire ten programmers
and deluge the hobby market with good software."
- Bill Gates, 1976, in want of www.gnuherds.org

[MJ Ray-2]

"Dave Crossland" <dave@...> wrote:
> 2008/8/30 Evan Prodromou <evan@...>: [...]
> > > B. Made publicly available."
>
> This should be "Made available to its users."
>
> Requiring publication is highly contentious.

Not really.  We don't have cost-free publication, so it's a cost to be
paid by the hosting user and so obviously any software requiring
publication is not free software. Even worse, if it has to be
published as long as the software is used, it's a near-unlimited cost.

> In the early "Emacs commune" days RMS decided it was unethical to
> force people to publish their changes, which is why the GPL doesn't
> (no link, I cuss printed books, I believe I read this in 'Free For
> All' or 'Rebel Code'). The eCos Public License that does what RMS once
> did, and the Reciprocal Public License that requires publication, is
> criticised at http://www.gnu.org/philosophy/license-list.html#NonFreeSoftwareLicense

I didn't find it in "Rebel Code".  I don't have "Free For All" here.

> More recently, debian-legal have the desert island/dissident tests [1]
> that, while not as important as the FSF FSD and OSI OSD, are still
> worth thinking about and should not be disregarded lightly.

They are both tests of Debian Free Software Guidelines which are also
covered in the FSD (DFSG's daddy) and the OSD (DFSG's bastard child).
Anything meeting FSD and OSD should meet DFSG and these tests by
definition, else one of them is broken and needs repair.

I last posted the links between the tests and DFSG in
http://lists.debian.org/debian-legal/2008/08/msg00123.html

> And most recently, the Affero GPLv3 says that the _users_ of network
> software must be able to access it via the network - which I think is
> striking the correct balance, although some d-l folks like MJ Ray
> don't like it.

The cost of publication is limited to the number of potential users,
but that means you can't use Affero GPLv3 software for a public
service unless you're prepared to pay for an unlimited number of
copies.

There's a bit of debate whether you can publish once to some free
hosting service and carry on using it even if the free hosting service
has gone down, but that's not obviously allowed, so I think AGPLv3 is
unacceptable, based on the utter nonsense that a licence can "ensure
cooperation" when cooperation as I understand it is necessarily
autonomous and voluntary as basic principles.
http://www.ica.coop/coop/principles.html

Now that FSF has tried to misrepresent cooperation, I see it is moving
on to back a campaign called autonomo.us which seems to be a handful
of people from the usual famous non-autonomous hierarchical
organisations.  Do those people have some experience of autonomous
action which has been hereto hidden from view?  Are we going to get
the pain of watching them learn from well-funded mistakes?

When will FSF return to free software and stop these ultra-vires
projects?  Even when its heart is in the right place, such as
DefectiveByDesign, some of its actions have been spectactular
own-goals, such as the Apple Store Helpdesk-clogging attack?  Has FSF
gone out of control?

> [1]: http://en.wikipedia.org/wiki/Debian_Free_Software_Guidelines#debian-legal_tests_for_DFSG_compliance
>
> My suggestion:
>
> An open software service is one:
>
> 1. Whose data is open as defined by the open knowledge definition
> (http://opendefinition.org/1.0/) with the exception that where the
> data is personal in nature the data need only be made available to the
> user (i.e. the owner of that account).
> 2. Whose source code is Free/Open Source as defined by both the FSF
> and OSI (that is available under a license in the OSI and the FSF
> approved lists) and is available to the users of the service.

I'd strike the ()s but other than that, it seems fine to me.

Regards,
--
MJ Ray (slef)
Webmaster for hire, statistician and online shop builder for a small
worker cooperative http://www.ttllp.co.uk/ http://mjr.towers.org.uk/
(Notice http://mjr.towers.org.uk/email.html) tel:[phone]

[davelab6]

2008/8/31 MJ Ray <mjr@...>:
>
>> In the early "Emacs commune" days RMS decided it was unethical to
>> force people to publish their changes, which is why the GPL doesn't
>> (no link, I cuss printed books, I believe I read this in 'Free For
>> All' or 'Rebel Code').
>
> I didn't find it in "Rebel Code".  I don't have "Free For All" here.

Ah, it was Free As In Freedom:

"Looking back, Stallman says the GPL compromise was fueled by his own
dissatisfaction with the Big Brother aspect of the original Emacs
Commune social contract. As much as he liked peering into other
hackers' systems, the knowledge that some future source-code
maintainer might use that power to ill effect forced him to temper the
GPL.

"It was wrong to require people to publish all changes," says
Stallman. "It was wrong to require them to be sent to one privileged
developer. That kind of centralization and privilege for one was not
consistent with a society in which all had equal rights."

- http://en.wikisource.org/wiki/Free_as_in_Freedom/Chapter_9

> I last posted the links between the tests and DFSG in
> http://lists.debian.org/debian-legal/2008/08/msg00123.html

Thanks! :-)

> The cost of publication is limited to the number of potential users,
> but that means you can't use Affero GPLv3 software for a public
> service unless you're prepared to pay for an unlimited number of
> copies.

Doesn't Bittorrent mitigate these costs?

> There's a bit of debate whether you can publish once to some free
> hosting service and carry on using it even if the free hosting service
> has gone down, but that's not obviously allowed,

But do you think publishing once to a free hosting service and carry
on using it while the free hosting service is up is allowed?

> the utter nonsense that a licence can "ensure
> cooperation" when cooperation as I understand it is necessarily
> autonomous and voluntary as basic principles.
> http://www.ica.coop/coop/principles.html

Does the GPL ensure co-operation for traditionally distributed software?

I think so.

> Now that FSF has tried to misrepresent cooperation, I see it is moving
> on to back a campaign called autonomo.us which seems to be a handful
> of people from the usual famous non-autonomous hierarchical
> organisations.

The about page lists the organisational affiliations as Novell, OKFN,
FSF, SFLC, CC, CivicActions, GNOME Foundation. But this is a handful
of people, and their affiliations seem totally secondary.

What autonomous heterarchical organisations do you have in mind that
may have members interested in joining that group?

> Do those people have some experience of autonomous
> action which has been hereto hidden from view?  Are we going to get
> the pain of watching them learn from well-funded mistakes?

None of them seem like microserf cubical dwellers to me;

> seems fine to me.

Great :-)

--
Regards,
Dave

[Thomas Lord]

Evan Prodromou wrote:

Hi, all. I'm interested in getting some feedback from the list on the Franklin Street Statement:

http://autonomo.us/2008/07/franklin-street-statement/

One of the recommendations to developers says:

Develop software that can replace centralized services and data storage with distributed software and data deployment, giving control back to users.

That is where the real essence of the problem lies.   It's the most important area in which to work.

Currently, web services are built out of conventional operating systems (often a LAMP stack or similar) and, at scale, web services are built out of custom-designed networks.

Consequently, currently, web services are a game for the rich and technologically savvy.   Only the rich and savvy can afford to and can figure out how to set up a typical web service.   Typical users, therefore, have no control (other than "take it or leave it") over what server-side code they run.

Contrast that with the *personal computer*.    Nearly any fool can buy a PC at the local CheapStuffMart and nearly any fool can figure out how to install that new poker-playing software or photo album organizer or whatever. 

Notice how that contrast relates to software freedom.   The concept of software freedom is *meaningful* in the context of personal computing because any fool can learn to study, modify, and share software -- it's an economically viable option for users to actually *exercise* software freedom, and so it is worth fighting for the right to do so.   In contrast, almost nobody can, as a practical matter, operate a web service:  it takes too much money to set up and too much expertise to duct tape together a network for it.   So EVEN IF all of the server-side code is openly shared free software, there is no practical way for the average fool to actually exercise software freedom over that code -- no meaningful way to use that freedom.   EVEN IF the code is free, users are still under the power of service operators.

The political cause of freedom for web service users is first and foremost an *engineering problem*:  we need a new server-side OS.

We need a server-side OS that allows each user to fully control the servers they use, including (and especially) allowing each user to change what programs are installed server-side.   Once we have that new OS, the question of software freedom for server-side code becomes meaningful and the question of user control over data becomes actionable.

One way to work on such an OS is to work on the problem of how to sell user-controlled services.   Amazon EC2 is an example of a project working on that one.

Another way to work on such an OS is to work on the problem of operations: keeping an application network functioning and scaling.   The RAD lab at Berkeley is an example of a project working on system administration monitoring and control systems in that area.

A third way to work on such an OS is to work on the platform specification -- the "API" to the OS.    Should it be some extension of Posix?  Should it be something else entirely?   What form will free software programs for this new OS take?   My own work (http://basiscraft.com) is an example of  working on that problem.

Franklin Street seems like a nice enough document but it lacks teeth and there's no urgency or sense of inevitability behind it.   Compare that to the early days of the free software movement:

The free software movement began with statements of principle, sure, but that is not what propelled it to the success it has had since.   Rather, it was the early days of the GNU project -- the engineering effort -- that gave the free software movement credibility and the problem of software freedom urgency.

The original releases (Emacs, GCC, and GDB -- all released in a short-time frame, followed fairly quickly by the shell utils, text utils, make, and autoconf) WERE, in essence, an OS.   They comprised a parasitic OS, unable to run on its own -- only on a host system.   Yet they were clearly a minimalist OS ready to grow into a complete system in fairly short order (as they did).   

Those early releases -- that "hacktivism" or "engineering activism" -- is what gave teeth to the free software movement.   They made the question of software freedom a *practical* question, faced by real users, rather than just a theoretical principle.

Perhaps Franklin Street needs its own version of a GNU project:  a vision for the new OS and an engineering plan for how to achieve it.



-t

[Stephen J. Turnbull]

Evan Prodromou writes:
 > Hi, all. I'm interested in getting some feedback from the list on the
 > Franklin Street Statement:
 >
 > http://autonomo.us/2008/07/franklin-street-statement/

I find it curious that they recommend that developers use Affero-style
licenses, but not for the services themselves.

 > ...and on the Open Software Service Definition:
 >
 > http://www.opendefinition.org/ossd

The "publicly available" part of the OSSD to be pushing too far.

In the Open Knowledge Definition, even pricing of distribution
services is regulated.  That is clearly over the boundary of freedom
and openness into socialism, and should be unnecessary given the no
royalty provision in clause 2.  Note that clause 2 permits
proprietization of the open knowledge.

[Russ Nelson]

MJ Ray writes:
 > Not really.  We don't have cost-free publication, so it's a cost to be
 > paid by the hosting user and so obviously any software requiring
 > publication is not free software. Even worse, if it has to be
 > published as long as the software is used, it's a near-unlimited cost.

Perhaps you're new here?  I've requested that people not use "free" on
this list to reduce confusion.  Instead, use ''gratis'' and ''libre'',
whose meaning is obvious in English, and literal in other
latin-derived languages.

We don't have cost-free publication of binaries either, so obviously
any software published in binary form is not gratis software.  But we
don't give a whit about gratis software.  It's libre software we care
about.  And what could be MORE freeing than being able to get the
source of code even if you haven't gotten the binary?

Remember: it's not about freedom for the developer.  It's about
freedom for the user of the software.  The power relationship granted
by copyright law gives all the freedoms to the developer.  Free
software is about the user insisting that the developer give up some
freedom -- in this case the freedom to hold back source code even from
people who haven't gotten the binary.  In exchange the user can, and
is expected to, help the developer create the software.  If there were
no benefits and only pain, who would be eager to write open source
software?  And yet everyone[1] is, so there must be a light under that
bushel.

[1] Even Microsoft.  Even Adobe.

 > The cost of publication is limited to the number of potential users,
 > but that means you can't use Affero GPLv3 software for a public
 > service unless you're prepared to pay for an unlimited number of
 > copies.

Bittorrent allows you to control the cost of that copying, however.
Buy hosting that allows X megabytes of download per day (which, for a
small value of X, is extremely cheap), then configure bittorrent to
stay under X.  That doesn't limit the number of copies (which I think
you would agree is improper), but it limits your cost.

--
--my blog is at    http://blog.russnelson.com   | Software that needs
Crynwr sells support for free software  | PGPok | documentation is software
521 Pleasant Valley Rd. | [phone]       | that needs repair.
Potsdam, NY 13676-3213  |     Sheepdog          |

[Stephen J. Turnbull]

Russ Nelson writes:

 > We don't have cost-free publication of binaries either, so obviously
 > any software published in binary form is not gratis software.  But we
 > don't give a whit about gratis software.  It's libre software we care
 > about.  And what could be MORE freeing than being able to get the
 > source of code even if you haven't gotten the binary?

 > Remember: it's not about freedom for the developer.

C'mon, Russ.  Be civil.  MJ is not one of those who needs reminding.

And you're wrong, anyway.  None of the definitions of software freedom
(whatever it's called) permit excluding people just because they're
software developers.  The question is, as always, when the exercise of
freedom by upstream users inhibits the freedom of downstream users,
who gets precedence?

The answer of the "permissive" school is "upstream: the only right we
take from those who choose non-libre distribution modes is the right
to use words like 'libre', 'open source', or 'free' to describe their
distribution; downstream can and should demand libre terms depending
on their needs."  The answer of the-school-I-like-a-lot-less is
"downstream: those users need to be forced to be free because they are
generally unwilling to pay the cost of negotiating a free distribution
mode with upstream."[1]

Licenses implementing *both* extremes impose externalities.  Strong
copyleft licenses take away my freedom to negotiate away my right to
source in return for a lower price, or even the very existence, of a
derivative work.  Permissive licenses, on the other hand, allow
intermediaries to refuse access to source of derivatives, and thus
restrict the freedom of downstream users who do care to get source.

I don't see an a priori liberal answer to this problem of the second
best, unless you're willing to go down Stallman's narcissistic path of
asserting that there is no freedom but software freedom that needs to
be considered here.

 > Bittorrent allows you to control the cost of that copying, however.
 > Buy hosting that allows X megabytes of download per day (which, for a
 > small value of X, is extremely cheap), then configure bittorrent to
 > stay under X.  That doesn't limit the number of copies (which I think
 > you would agree is improper), but it limits your cost.

You will probably argue that users (ie, those who want to download via
bittorrent) are free to find better ISPs, but in my experience ISPs
often interfere with bittorrent.  Bittorrent also sucks for long-term
support, as it can take forever to download if all you've got out
there is the original seed.  This is a PR liability that must be
borne.  There's also the permanent legal/admin cost of keeping track
of all these specific rules.  I don't think it's fair to just pooh-
pooh the costs proposed to impose on others, when a careful study of
what those costs might be has not been done.

Footnotes:
[1]  Note that I do not object to copyleft licensing, or Affero
licensing, if that's what the owner of the work wants.  I just don't
consider those licenses "better" from the viewpoint of enhancing
freedom than the permissive ones.

[Russ Nelson]

Stephen J. Turnbull writes:
 > Russ Nelson writes:
 >
 >  > We don't have cost-free publication of binaries either, so obviously
 >  > any software published in binary form is not gratis software.  But we
 >  > don't give a whit about gratis software.  It's libre software we care
 >  > about.  And what could be MORE freeing than being able to get the
 >  > source of code even if you haven't gotten the binary?
 >
 >  > Remember: it's not about freedom for the developer.
 >
 > C'mon, Russ.  Be civil.  MJ is not one of those who needs reminding.

Oh dear.  That WAS the civil version.

--
--my blog is at    http://blog.russnelson.com   | Software that needs
Crynwr sells support for free software  | PGPok | documentation is software
521 Pleasant Valley Rd. | [phone]       | that needs repair.
Potsdam, NY 13676-3213  |     Sheepdog          |

[Stephen J. Turnbull]

Russ Nelson writes:
 > Stephen J. Turnbull writes:

 >  > C'mon, Russ.  Be civil.  MJ is not one of those who needs reminding.
 >
 > Oh dear.  That WAS the civil version.

I'm curious, what did the CRIMINAL version look like?

Seriously, this is obviously a long-standing conversation between you
guys, but for the reasons I gave, I don't see why you (of all the
libre advocates I know) are going down that road.  *Advocate* such
practices or licenses as freedom-enhancing, that's plausible.  But
requiring extreme conditions (such as publication) in the *definition*
of openness seems paradoxical to me.

[davelab6]

2008/9/2 Stephen J. Turnbull <stephen@...>:
>
> requiring extreme conditions (such as publication) in the *definition*
> of openness seems paradoxical to me.

Which is why requiring users have access to source code, even if its a
network application so the binary isn't on their machine, sounds good
to me :-)

Best,
Dave

[Russ Nelson]

Stephen J. Turnbull writes:
 > Seriously, this is obviously a long-standing conversation between you
 > guys,

No, it's not.  I had no idea that anybody would say that because
software distribution (of either source or binaries) cannot be done
without cost, that libre software does not exist.

--
--my blog is at    http://blog.russnelson.com   | Software that needs
Crynwr sells support for free software  | PGPok | documentation is software
521 Pleasant Valley Rd. | [phone]       | that needs repair.
Potsdam, NY 13676-3213  |     Sheepdog          |

[Stephen J. Turnbull]

Russ Nelson writes:
 > Stephen J. Turnbull writes:
 >  > Seriously, this is obviously a long-standing conversation between you
 >  > guys,
 >
 > No, it's not.  I had no idea that anybody would say that because
 > software distribution (of either source or binaries) cannot be done
 > without cost, that libre software does not exist.

I still don't see that he said *that*.  I see that he's saying that
the open service definition can impose excessive costs on developers,
in addition to kneecapping their revenue possibilities (relative to
non-libre distribution, not relative to some notion of "fair
compensation").  And he doesn't want those costs imposed.

There are also the desert island and dissident tests which pretty much
exclude publication requirements.  I think those are a sine qua non if
you are a liberal advocate of libre software rather than a socialist
antagonist of private property.

After some thought, I don't find his arguments (which I may not
understand, as they are mostly implied) as plausible as I originally
did.  I myself still object to "publication as open source" as a
requirement.  However, bandwidth etc are just costs of business, and
(once an Affero-like clause is in place) no different from the
existing GPL requirement that allows choice of immediate distribution
of source or a written offer to distribute (which is actually somewhat
more strict since it can extend up to 3 years after you stop
distributing the binary).