9 Jul
2019
9 Jul
'19
11:50 a.m.
Dear Osmocom community,
during recent years we have developed a quite extensive set of test suites
for virtually all interfaces on all protocol levels of our 2G stack, and
recently even extended at least with some 3G IuCS coverage.
Those test suites are considerably valuable for testing any kind of
implementation of 2G / 3G network elements, even beyond Osmocom. I'm a
bit worried that they might be used to test proprietary implementations,
which would bring no value to growing FOSS in mobile communications.
Whether we license the test suites under GPL or AGPL doesn't actually
change much here: Imagine "EvilCorp" developing a BSC, using our test
suites against it. They neither distribute the test suite, nor do they
provide "access over a network" to it, an hence they could keep any
modifications/extensions/derivatives private without having to
contribute those back.
I think this is problematic. We develop those test suites because we
care about having well-tested FOSS in cellular communications, whether
Osmocom or other FOSS projects. I certainly don't want to spend my
spare time, or invest sysmocom resources towards improving the test
coverage of any non-FOSS implementations. If such users are out there,
they should at the very least contribute to the development effort in
one way (code) or another (funding).
For 2G/3G, I think this discussion is quite theoretical, as there is
unlikely anyone else implementing those old technologies in proprietary
systems - Osmocom is probably typically the latest player in the market
to do so.
However, as I'm currently working on a first set of TTCN3 tests for
testing a LTE MME (initially attacing to S1 and SGs), I'm wondering if
we should continue to release related code under traditional copyleft
licenses.
One idea would be to have a license that permits the use of the test
suites to test only FOSS software. In theory, that would mean that all
projects we care about (such as currently srsLTE, nextepc, OAI EPC, ...)
could use the test suites to test their software. On the other hand,
if some vendor of proprietary MME/EPC software would want to [legally]
use the tests, they would need to obtain a different (paid) license to
the test suite. Of course they could simply ignore the license and we'd
have little chance to learn about it - but I would argue most proper
companies typically would obtain a license if they used some software
they knew they had to license.
This is a very double-edged sword. Drafting such a new license would
cause license proliferation, which is always bad. It also raises
quesetions of license compatibility with e.g. some of our common
existing 'library' code that would need to be investigated. Finally, it
also means that we'd no longer be writing "free software" nor "open
source", as the respective relevant definitions always require "non
discriminatory use for any purpose", which would no longer be the case
here.
This is currently an early stage brainstorming. Now is the right point
in time to talk about this, before we release any LTE/EPC related tests.
Let me know if you have any thoughts to share. While I'm cross-posting
this to openbsc@ and nextepc@ lists, pleaes follow-up-to
openbsc(a)lists.osmocom.org, as that's the main Osmocom mailing list and
we don't have any specifically only for the test suites.
--
- Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)
10 Jul
10 Jul
8:54 a.m.
On Tue, Jul 09, 2019 at 07:50:26PM +0800, Harald Welte wrote:
We develop those test suites because we
care about having well-tested FOSS in cellular communications, whether
Osmocom or other FOSS projects. I certainly don't want to spend my
spare time, or invest sysmocom resources towards improving the test
coverage of any non-FOSS implementations. If such users are out there,
they should at the very least contribute to the development effort in
one way (code) or another (funding).
I understand your concern. But FOSS often ends up being used in places
and ways which the original authors of the software don't endorse for
moral or ethical reasons. This problem exists for all FOSS projects in
some form or another. There could be some useful "prior art"; have you
already looked for similar discussions elsewhere?
Finally, it
also means that we'd no longer be writing "free software" nor "open
source", as the respective relevant definitions always require "non
discriminatory use for any purpose", which would no longer be the case
here.
Indeed. The best way to limit downstream consumers to a subset of "everyone"
would be a proprietary licence. I don't think Osmocom should adopt one.
The usual approach would be to accept the risks of inadvertently helping
proprietary competitors, stick to a well-known FOSS licence, and hope that
the benefits outweigh the risks in the long term.
11:47 a.m.
Hi Stefan,
On Wed, Jul 10, 2019 at 10:54:27AM +0200, Stefan Sperling wrote:
Certainly I would never propose to use a non-FOSS license for the actual
Osmocom software, i.e. the 3GPP protocol and functional elements
implementations we're working on. Osmocom *is* about open source
cellular implementations, and nothing else.
But I think for test suites it is something worth a debate. I actually
started to have related thoughts already about osmo-gsm-tester and the
existing TTCN-3 test suites. Those kind of test suites are rarely ever
executed by a user. So it isn't relevant to the freedom of the end user
(typically a small commercial or non-commercial operator) if the actual
test suites are FOSS or not. Availability of the test suites matters
most to developers, and I would argue that it mostly is about 'source
available', and not about 'open source'.
On Tue, Jul 09, 2019 at 07:50:26PM +0800, Harald Welte
wrote:
I agree. However, I still think that a test suite for "protocol/spec
conformance" to an industry standard is something different here.
1) the test suite is not the main "product" of Osmocom, but it is merely
a tool to test our FOSS software. Most projects don't face the
problem that way, as most unit tests or other tests invariably cover
very specific aspects of their software (APIs, data structures,
custom interfaces of any sort), and hence the usefulness of those
tests beyond the project is very limited.
2) I believe the fundamental reason why there's a lack of published test
suites for most protocol specs out there is that all respective
developing companies/entities keep them private. There are even
examples of companies developing [dual licensed] FOSS software but
keeping the test suites private. The only examples that come to mind
in the cellular sphere are the ETSI conformance test suites published
in TTCN-3, but mostly in a format that cannot be consumed directly by
TITAN (or any other TTCN3 compiler/runtime) as larte parts about
interfacing the test system are left for the user to implement. And
they have restrictive/unclear licenses. But even outside the 3GPP
world, I think there aren't many [FOSS] conformance/compliance tests out
there for protocols. I don't recall having seen any of those for
the many protocols I worked with in the past, e.g.
FTP/SMTP/POP3/IMAP/NNTP/PPTP - so the IETF world doesn't appear to be
much different.
We develop those test suites because we
care about having well-tested FOSS in cellular communications, whether
Osmocom or other FOSS projects. I certainly don't want to spend my
spare time, or invest sysmocom resources towards improving the test
coverage of any non-FOSS implementations. If such users are out there,
they should at the very least contribute to the development effort in
one way (code) or another (funding).
I understand your concern. But FOSS often ends up being used in places
and ways which the original authors of the software don't endorse for
moral or ethical reasons. This problem exists for all FOSS projects in
some form or another. There could be some useful "prior art"; have
you
already looked for similar discussions elsewhere?
I'm not aware of a discussion specifically for protocol
functional/conformance testing. There are of course the various
discussions about licenses in the "cloud world" where even AGPL is
deemed insufficient by some entities. I don't like that development in
general, as
a) those entities call their software still 'open source' when it should
instead be 'source available'
b) the licenses are for the actual 'product' and not for a test suite
around that product
Finally, it
also means that we'd no longer be writing "free
software" nor "open source", as the respective relevant definitions
always require "non discriminatory use for any purpose", which would
no longer be the case here.
Indeed. The best way to limit downstream consumers to a subset of "everyone"
would be a proprietary licence. I don't think Osmocom should adopt one. The usual approach would be to accept the risks of
inadvertently helping
proprietary competitors, stick to a well-known FOSS licence, and hope that
the benefits outweigh the risks in the long term.
Osmocom is a small speckle of dust in the world of cellular
communications, while virtually everyone, even the very smallest
proprietary player has more significance in terms of market adoption
as well as 'revenue' surrounding the project/product compared to
proprietary vendors. (I'm talking about 'pure' proprietary vendors
here, not the likes of Range Networks, SRS, ...)
We're working on a seemingly insurmountable task in terms of both breadth
(all network elements of from PHY to CN in 2G/2.5/2.75G, CN for 3G/3.5G,
and now 4G) as well as complexity of the related technology, with
incredibly limited resources and a surprisingly small team. We have to
make sure that the result of our work counts, and that those limited
resources are invested very wisely.
The amount of contributions outside of the relatively small and dedicated
community of developers and users has been small (but existant!) for the
actual Osmocom CNI programs, particularly in recent years. But looking
at the osmo-ttcn3-hacks repository, I cannot find a single commit in the
tree that was not written either by somebody on sysmocom payroll or by
Vadim (yay! thanks!).
Meanwhile, the business model of the tradiditonal cellular industry is
very much anti-FOSS, if not only due to the fact that a (if not the)
largest revenue comes from licensing "essential patents" on what is
written in the 3GPP specs. So I would not expect any significant change
in terms of contributions to related projects from the traditional
proprietary industry even in the forseeable future.
In the end, I think that proprietary vendors interested in the test
suites would actually not have a problem licensinge it under
proprietary/commercial terms - this is their normal business
irrespective of where they source that testsuite from. Any related
resources help us to co-fund the development of the tests - and
FOSS projects can still use them.
Regards,
Harald
--
- Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)
12:48 p.m.
Hi Harald, Stefan.
I'd just chime in here to say that this is something I've been thinking
about recently and trying to discuss when I get a chance.
The topic of said discussion is whether the time may have come to revise
the FOSS "ethic";
Is there just too much "evilcorp" out there now to justify continuing to
hand them our labour on a plate? - A reference I sometimes use here is
the linux kernel as a base for a certain "evil" surplus collecting and
behaviour modifying robot that was first hawked to the world as a
"liberating" "open-source" project.
Once it had it's evil robot paw on our hearts and minds and those of our
children, it somehow felt the need to become somewhat less open. While I
myself have never added a single line of code or even submitted a bug
report on the linux kernel, I think I would feel somewhat strange about
the above if I had.
Of course, one could point to a lot of projects that retroactively
benefit from open development communities and developers in the pay of
some evilcorp or other have indeed contributed a lot of FOSS!
Still, evilcorp will work in evilcorp's interests and nobody else's.
Yes, there's a type mismatch in that sentence, ( type corp cannot be
compared with type person ), but the point is that as long as FOSS is
good for corp, corp will support it, but corp is getting VERY big these
days and so the days for FOSS (or maybe I should say FOSS that is not in
corp's interest) might be numbered.
So for what it's worth, in principal I'm with Harald, especially given
that I recognise that if I am to continue development on osmocom, I need
to start contributing to the test suites!
I have NO IDEA about the actual licensing issues or what the eventual
license would/should look like. Luckily there are people around here who
are quite expert in that field.
K.
2217
days inactive
2218
days old
3 comments
3 participants
participants (3)
-
Harald Welte -
Keith -
Stefan Sperling