Real men don't attack straw men

[Reid Nichol]

--- Richard Stallman <rms@...> wrote:
>     Thus the combined work, THE WHOLE POINT OF WRITING IT, is under
>     the GPL.  That IS what you just said.  Which is forcing me into a
>     license for my project that I don't want.
>
> We require you to use, for your program that contains our code,
> a license that protects the essential freedom for all its users.
> That defends real freedom.


You mean your twisted definition of freedom.  Btw, your own FAQ states
that I can't BSD my code if I link to a GPL'd lib.  Contrary to what
you said I might add.  I think you need to read your own FAQ.

http://www.gnu.org/licenses/gpl-faq.html

And find out what freedom actually means:

http://dictionary.reference.com/browse/freedom

I would comment further, and on other things, but I believe that you're
too far gone to warrant any more time spent on this.  At least from me
and as it seems others as well.  That is, until you gain some sanity.


best regards,
Reid Nichol

President Bush says:

War Is Peace
Freedom Is Slavery
Ignorance Is Strength


      ____________________________________________________________________________________
Never miss a thing.  Make Yahoo your home page.
http://www.yahoo.com/r/hs

[Alexander Terekhov-3]

On Jan 12, 2008 1:49 AM, Reid Nichol <rnichol_rrc@...> wrote:
> --- Richard Stallman <rms@...> wrote:
> >     Thus the combined work, THE WHOLE POINT OF WRITING IT, is under
> >     the GPL.  That IS what you just said.  Which is forcing me into a
> >     license for my project that I don't want.
> >
> > We require you to use, for your program that contains our code,
> > a license that protects the essential freedom for all its users.
> > That defends real freedom.

Stallmanism cult wants you to believe in something that is "true" only
in the GNU Republic. Uncopyrightable aggregations[1] aside for a
moment, you, as a sole author of a compilation (this term includes
collective works), do have all the rights in compilation work. That's
one difference (among others) between compilations and derivative
works. Your compilation copyright is totally independent from
copyrights on constituent works.

But in the GNU Republic, the copyr^Hleft act has created fascinatingly
fuzzy regime for software ("quanta" mismatch and all that, see below).
 It's not about expression (as in literary works per Berne Convention
which says that computer program works are to be protected as literary
works) modulo the AFC test[2] (to filter out unprotectable elements)
like in the rest of the world. Rather, as Eben The dotCommunist
Manifesto Moglen has nicely put it (in slight disagreement with RMS):

http://www.fsfeurope.org/projects/gplv3/bangalore-rms-transcript

-----
Q10c: Lets say I have a program that uses free libraries, which are...

Richard Stallman: Well, linking them together like that is clearly
combining them. The rules, based on the existing GPL, are too
complicated for me to try to recite them to you. All I can say is, yes,
the GPL makes conditions in that case.

Q10d: That means any such use is a violation of the GPL?

Richard Stallman: Some kinds may be permitted. That's why I'm saying it
depends on details, very much. But linking components together is
certainly combining them.

Eben Moglen: Richard, can I make a comment here? Here's the problem.
The problem that you're facing in asking the question, and the problem
that Richard is facing in trying to answer it. When you try to take two
disciplines of thought that use different primitive quanta - different
units of meaning - there's not going to be a congruent mapping between
one vocabulary and the other - as there is no guarantee that there is a
one-to-one match between words in Hindi and words in English.

The problem is that the unit of meaning in copyright law is the work,
whatever the work is. That's the unit in which copyright law speaks. So
the author, or authors, of a work have certain exclusive rights,
including the rights to control modification and distribution. GPL
says, we give most of those rights to the user, in the work, rather
than withholding them, as proprietary users do.

What's the unit of a program? Not the work. Computer science has
defined many quanta of meaning in computer program since I began
decades ago. The subroutine, the function, the module, the object. Each
of those is a unit of meaning in a language of computer activity, but
it's not the work under copyright law.

Between the the quantum: work, and the quantum: module, library, file,
function, object, procedure, there is not a one-to-one mapping, and the
consequence is that when we attempt to exert our intention in copyright
law, we only speak in terms of the work. We must use the vocabulary of
copyright. Since that doesn't map neatly to the vocabulary of computer
programming, no matter what that vocabulary happens to be, given the
dominant paradigm of program construction, there is guaranteed to be a
zone of uncertainty.

Richard Stallman: I disagree. I wouldn't say that you're wrong. What
you're saying is right, but there's something even deeper to be said,
which is that what you're saying is not a problem. It sounds like
you're describing a problem, but in fact, criteria... because of the
fact that in a program you can express the same thing in many different
ways, and you can rewrite it to use many different ways to communicate,
any kind of criteria drawn up in terms of the technical boundaries that
exist in programs would be a bad criterion because it would be too easy
to play games with it.

If there were a criterion about files, well, it's easy to move
something from one file to another. If the criteria were about
subroutines, it's easy to split up a subroutine. You see what I mean?
Any criteria formulated in terms of the technical entities of
programming would be too easy to game around.

Eben Moglen: As when, for example, people tried to draw a line between
static linking and dynamic linking under GPL version two, and we had to
keep telling people that whatever the boundary of the work is under
copyright law, it doesn't depend upon whether resolution occurs at link
time or run time. Right? Those kinds of technical decisions, whatever
they are, don't map neatly into the language of copyright, which is the
language of the licence.

Richard Stallman: Nor into the intentions of the GPL. Because, the
point is, if we drew the line in the kind of clear way that programmers
want, in terms of technical points, then it would be easy for somebody
to evade the intention of the GPL just by taking that line as the
instructions on how to do it.

So, by making it so clear, in a mechanical sense, we would be
undermining the goal.
-----

>
>
> You mean your twisted definition of freedom.  Btw, your own FAQ states
> that I can't BSD my code if I link to a GPL'd lib.  Contrary to what
> you said I might add.  I think you need to read your own FAQ.
>
> http://www.gnu.org/licenses/gpl-faq.html
>
> And find out what freedom actually means:
>
> http://dictionary.reference.com/browse/freedom

[1] http://www.usfca.edu/law/determann/softwarecombinations060403.pdf

-----
DANGEROUS LIAISONS  SOFTWARE COMBINATIONS AS DERIVATIVE WORKS?

Distribution, Installation and Execution of Linked Programs under
Copyright Law, Commercial Licenses and the GPL

By Lothar Determann

Prof. Dr. Lothar Determann teaches courses on Computer and Internet
law at the University of California Berkeley School of Law (Boalt Hall),
University of San Francisco School of Law and Freie Universitdt Berlin
(www.lothar.determann.name) and practices law as a partner in the
international technology practice group of Baker & McKenzie LLP, San
Francisco/Palo Alto office (www.bakernet.com). The author is grateful
for assistance from his students, in particular Tal Lavian, Principal
Scientist at Nortel Labs (valuable comments from computer science
perspective), Steven B. Toeniskoetter, Lars F. Brauer, and Neda
Shabahang (legal research and footnote editing).
-----

[2] http://www.digital-law-online.com/lpdi1.0/treatise24.html
(IV. Applying The AFC Test)

See also

http://digital-law-online.info/lpdi1.0/treatise27.html
(VI.D.4. Derivative Works and Compilations)

regards,
alexander.

P.S. HOUSE REPORT NO. 94-1476:

http://en.wikisource.org/wiki/Index:H.R._Rep._No._94-1476

-----
Between them the terms ''compilations'' and ''derivative works''
which are defined in section 101 comprehend every copyrightable
work that employs preexisting material or data of any kind. There
is necessarily some overlapping between the two, but they basically
represent different concepts. A ''compilation'' results from a
process of selecting, bringing together, organizing, and arranging
previously existing material of all kinds, regardless of whether
the individual items in the material have been or ever could have
been subject to copyright [...] an unauthorized translation of a
novel [i.e. derivative work] could not be copyrighted at all, but
the owner of copyright in an anthology of poetry [i.e. compilation]
could sue someone who infringed the whole anthology, even though
the infringer proves that publication of one of the poems was
unauthorized.
-----

[Andrés Delfino]

1. Stallman states that Linux current version is partially non-free. *1

A program can't be partially non-free. A program is free if users have
the four freedoms, if not, it is non-free. The users of Linux does not
have the freedom to access the source code of parts of it (freedom 1).

2. Stallman states that Torvald's version of Linux is non-free. *2

There is no free version of Linux. Free operative systems which use
modified versions of Linux exist, but since those modifications are
not adopted by Torvalds's project (Linux), the term GNU/Linux can't
refer to a free version of it.

Even modifying the meaning of the name Linux to refeer to any of those
free operative systems doesn't work, since Torvald's "version" of
Linux surpass by far the number of users that any such operative
system has. So the term GNU/Linux is not even practically "OK".

3. Stallman states that the name of Gobuntu (a GNU/Linux operative
system more free that any popular GNU/Linux operative system) is so
close to Ubuntu that, practically speaking, it is not feasible to
recommend Gobuntu without recommending Ubuntu. *3

GNU/Linux is the preferred term for referring to a free operative
system for the Free Software Foundation and the GNU Project. Yet they
find OK the fact that they are using the name of a non-free kernel.

4. Stallman states that GNU developers didn't develop GNU just to make
it a technical triumph, or just to have a success. Their goal was to
win freedom, for they and for us. *4

They failed the moment they considered Linux, a non-free program,
fitted well as the kernel of its project; sufficiently well to not do
any serious development on its own kernel.

Hurd offers the GNU Project the option to release the much awaited GNU
operative system now. If winning freedom is the goal, then an inmature
free kernel is not a problem. And, ironically, the immature state of
Hurd is because of Linux. Active and continuous development on Hurd
does not ocurr because it isn't needed by the GNU Project. If Linux
hadn't been used in the first place by the GNU Project and the Free
Software Foundation, Hurd would be quite mature after 18 years of its
birth, and the free operative system envisioned by Stallman back in
1983 would be a reality.

5. The foundation directed by Stallman removes Linux from its Free
Software Directory but does not disassociate from it.




Be a real man, Richard, replace Linux with Hurd, fork Linux, or stop
the hypocrisy.

Either way, answer publicly, and not off-list.




*1
http://www.gnu.org/philosophy/linux-gnu-freedom.html

*2
http://article.gmane.org/gmane.os.openbsd.misc/134377

*3
http://article.gmane.org/gmane.os.openbsd.misc/134522

*4
http://www.pcworld.idg.com.au/index.php/id;211669437;pp;2

[Max Hayden Chiz]

Brian,

After your post (and several others), I tried BitTorrent out on my
network (sparc64 router + DOCSIS 2.0 cable connection; see
http://marc.info/?l=openbsd-misc&m=120019379210857&w=2)

After some experimentation, I was able to determine that running
BitTorrent with a large number of connections causes a huge increase
in latency regardless of bandwidth.  No one seems to know why this is,
but that might just be because my thread got buried by trolls and
other posts.

I'm not having watchdog timeouts but there is an off chance that the
latency increase that I experience and your timeout problem may be
related.  My work around is to use the max-src-states feature of pf to
limit the number of bit torrent connections to a reasonable number (50
seems to be a good trade-off on my machine, YMMV).

Could you modify your pf.conf to do this (or limit your connections at
the client and use pf to confirm) and let us know if that works on
your end as well?

--MHC

On Jan 5, 2008 1:22 PM, Brian <bwaichu@...> wrote:

> Is there any suggested PF setup when using BitTorrent?
>
> Right now, the biggest problem I have when using BitTorrent is watchdog
> timeouts.
>
> Thanks,
>
> Brian
>
>
>
>
>       ____________________________________________________________________________________
> Never miss a thing.  Make Yahoo your home page.
> http://www.yahoo.com/r/hs

[Chris Kuethe]

On Jan 14, 2008 4:06 PM, Max Hayden Chiz <max.chiz@...> wrote:

> Brian,
>
> After your post (and several others), I tried BitTorrent out on my
> network (sparc64 router + DOCSIS 2.0 cable connection; see
> http://marc.info/?l=openbsd-misc&m=120019379210857&w=2)
>
> After some experimentation, I was able to determine that running
> BitTorrent with a large number of connections causes a huge increase
> in latency regardless of bandwidth.  No one seems to know why this is,
> but that might just be because my thread got buried by trolls and
> other posts.

My theory is that you're using a ... uh... well, not very good
connection that bogs down easily. Some time ago, I tossed together a
little undeadly article on how to use altq to keep bittorrent from
dragging your network down -
http://www.undeadly.org/cgi?action=article&sid=20061109202501

I think you can solve this by tuning your service classes a little
better. I'm currently stuck on a 1.5M/384k dsl - to minimize the pain,
I told altq that i have 1.3M/340k.

CK

--
GDB has a 'break' feature; why doesn't it have 'fix' too?

[Max Hayden Chiz]

On Jan 14, 2008 6:30 PM, Chris Kuethe <chris.kuethe@...> wrote:
> My theory is that you're using a ... uh... well, not very good
> connection that bogs down easily.

My connection normally works fine; even when I max out my 7Mb/512Kb
line.  Running BitTorrent (even with a fraction of the bandwidth)
makes my latency go through the roof.

> Some time ago, I tossed together a
> little undeadly article on how to use altq to keep bittorrent from
> dragging your network down -
> http://www.undeadly.org/cgi?action=article&sid=20061109202501

See the thread I referenced in my previous email.  The issue I am
experiencing seems to have nothing to do with bandwidth usage.

> I think you can solve this by tuning your service classes a little
> better.

If you have a specific suggestion, I will try it and post the result
here, but as I discussed in the previous thread, altq shows that there
is no backlog of packets, and even with BitTorrent rate limited to a
small fraction of the bandwidth (via CBQ or HFSC) it is still able to
cause the latency issue.  By contrast if I limit the number of
connections, BitTorrent can consume almost all of the bandwidth and
the issue will not appear.

Perhaps this problem is specific to my configuration (or specific to
DOCSIS cable modems).  But if it makes Brian (or someone else's
problem) go away, then it is likely that this problem is not unique.

--MHC

[Brian-79]

--- Max Hayden Chiz <max.chiz@...> wrote:

> Perhaps this problem is specific to my configuration (or specific to
> DOCSIS cable modems).  But if it makes Brian (or someone else's
> problem) go away, then it is likely that this problem is not unique.
>
> --MHC
>

Let me read through the documentation to figure out how to set this up.  I am
running a cable modem as well.

Here are my bittorrent settings:

--minport 13000 --maxport 14000 --max_initiate 15 --max_allow_in 15
--max_upload_rate 25 --max_uploads 5

Give me some time to figure out the altq and pf.  I have only used pf for a
week, so I'm still learning it.

Thanks,

Brian



      ____________________________________________________________________________________
Looking for last minute shopping deals?  
Find them fast with Yahoo! Search.  http://tools.search.yahoo.com/newsearch/category.php?category=shopping

[Chris Kuethe]

On Jan 14, 2008 5:00 PM, Max Hayden Chiz <max.chiz@...> wrote:
> cause the latency issue.  By contrast if I limit the number of
> connections, BitTorrent can consume almost all of the bandwidth and
> the issue will not appear.
>
> Perhaps this problem is specific to my configuration (or specific to
> DOCSIS cable modems).  But if it makes Brian (or someone else's
> problem) go away, then it is likely that this problem is not unique.

i doubt it's your machine not being happy with number of connections -
i routinely have hundreds of states. depends on your modem, maybe? or
who made the board inside your modems? or what crack-addled rhesus
monkey pretended to write the firmware. If several different
manufacturers licensed the IP stack or NAT engine from the same
vendor, then it's perfectly possible that you both have ill-designed
hardware.

if you have a shell host somewhere, you could use netcat to determine
where your router falls over. use netcat to send /dev/zero over the
wire, with just a single stream, that should be pretty fast. then
start opening up more tcp connections (but don't send anything through
them) until it starts to bog down. That should give you enough
information to create several different rule classes to limit number
of connections by port. Web,ssh, dns and chat get to open connections
first, and can always open one more connection, all other ports get to
share ... 80% of max connections?

--
GDB has a 'break' feature; why doesn't it have 'fix' too?

[Johan Beisser]

On Jan 14, 2008, at 5:10 PM, Brian wrote:

> --- Max Hayden Chiz <max.chiz@...> wrote:
>
>> Perhaps this problem is specific to my configuration (or specific to
>> DOCSIS cable modems).  But if it makes Brian (or someone else's
>> problem) go away, then it is likely that this problem is not unique.

It's not unique, I saw the same issue recently. I basically exceeded  
the number of states my CPU/RAM combo could handle easily (roughly  
2400, "normal" average is 200 state rules) while pushing major amounts  
of data.

If I reduced the number of connections through bittorrent, performance  
improved.  During the download, at 350 peers, regardless of the  
download rate, I had 2400 some odd state rules. I suddenly saw round  
trip ICMP echo taking 900+ ms to the first hop. At 325, times were  
merely 90ms to the first hop, and normal is around 10ms.

The hardware is a slightly loaded Soekris net4501 with 64mb of RAM  
running OpenBSD 4.1 (GENERIC). It does not have a crypto accelerator,  
and handles ssh and openvpn on the main CPU (both are fairly low in  
usage at this time).

My guess, so far, is an issue with my ruleset, the hardware, and the  
use of synproxy for some of the TCP states (almost all of the BT  
clients I had were over TCP). OpenBSD itself seems to be fine, up  
until I get close to the limits of the hardware.

> Let me read through the documentation to figure out how to set this  
> up.  I am
> running a cable modem as well.
>
> Here are my bittorrent settings:
>
> --minport 13000 --maxport 14000 --max_initiate 15 --max_allow_in 15
> --max_upload_rate 25 --max_uploads 5
>
> Give me some time to figure out the altq and pf.  I have only used  
> pf for a
> week, so I'm still learning it.

Ask around if you have questions. There are excellent articles and  
examples available.

[knitti]

On 1/15/08, Chris Kuethe <chris.kuethe@...> wrote:
> i doubt it's your machine not being happy with number of connections -
> i routinely have hundreds of states. depends on your modem, maybe? or
> who made the board inside your modems? or what crack-addled rhesus
> monkey pretended to write the firmware. If several different
> manufacturers licensed the IP stack or NAT engine from the same
> vendor, then it's perfectly possible that you both have ill-designed
> hardware.

add me to the crowd. I got curious and tested a bit around and am able
up my ping latency up to 9 s (then I got bored :-). I *think* (at least here)
incoming connections are more of an issue than outgoing ones, and
traffic shaping (as in bandwith limiting) doesn't help much. Not traffic
shaping doesn't help either. This is on a 3M/512k ADSL
(Arcor/Germany) with a soekris and pppoe. I have not yet looked if
its just a pps issue with my soekris (in that case there should be no
latency problem at about 1/20th the available bandwith.)

I could imagine that some equipment not up to the task at the provider
could keep tabs on my states, but I don't see any reason why. (Yes I
know, there's this new evil data retention law, but the providers don't
even know what exactly they have to log and they are not exactly keen
on implementing it).

--knitti

[Stuart Henderson]

On 2008/01/14 19:40, johan beisser wrote:
> The hardware is a slightly loaded Soekris net4501 with 64mb of RAM
> running OpenBSD 4.1 (GENERIC).

This will handle much more traffic if you upgrade to 4.2.

[Johan Beisser]

On Jan 15, 2008, at 1:35 AM, Stuart Henderson wrote:

> On 2008/01/14 19:40, johan beisser wrote:
>> The hardware is a slightly loaded Soekris net4501 with 64mb of RAM
>> running OpenBSD 4.1 (GENERIC).
>
> This will handle much more traffic if you upgrade to 4.2.

I thought the performance improvement came from 4.1 with the removal  
of per packet interrupts.

The closest relevant information from plus42.html:
* Enable interrupt holdoff on DP83816 sis(4) chips. Significantly  
improves performance of such devices under load.

and from my dmesg:
sis0 at pci0 dev 18 function 0 "NS DP83815 10/100" rev 0x00, DP83815D:  
irq 10, address 00:00:24:c0:31:c8

So, I'm not entirely sure what you mean.

[Stuart Henderson]

On 2008/01/15 09:13, johan beisser wrote:

>
> On Jan 15, 2008, at 1:35 AM, Stuart Henderson wrote:
>
>> On 2008/01/14 19:40, johan beisser wrote:
>>> The hardware is a slightly loaded Soekris net4501 with 64mb of RAM
>>> running OpenBSD 4.1 (GENERIC).
>>
>> This will handle much more traffic if you upgrade to 4.2.
>
> I thought the performance improvement came from 4.1 with the removal of per
> packet interrupts.

http://www.openbsd.org/42.html
Huge performance improvements in the network stack, including:
# In pf, store routing table ID, queue ID etc directly in the packet
header mbuf instead of using mbuf tags (which use malloc'd memory). This
yields a 100% improvement in pf performance.
# Packet forwarding can skip IPSEC stack if no IPSEC flows are defined.
This yields a further 5% improvement in packet forwarding performance.
# Skip TCP/UDP/ICMP/ICMP6 checksumming when not necessary. This yields a
further 10% improvement in pf performance.

> The closest relevant information from plus42.html:
> * Enable interrupt holdoff on DP83816 sis(4) chips. Significantly improves
> performance of such devices under load.

that doesn't help your 83815D.

[Maurice Janssen-2]

On Tuesday, January 15, 2008 at 09:13:02 -0800, johan beisser wrote:

>On Jan 15, 2008, at 1:35 AM, Stuart Henderson wrote:
>>This will handle much more traffic if you upgrade to 4.2.
>
>I thought the performance improvement came from 4.1 with the removal  
>of per packet interrupts.
>
>The closest relevant information from plus42.html:
>* Enable interrupt holdoff on DP83816 sis(4) chips. Significantly  
>improves performance of such devices under load.
>

What about these:
- Improvement in the memory pool handling code, removing time from the
  pool header leads to better packet rates.
- Another pf(4) speed improvement. Delay packet checksumming in the PF
  code until we are sure we need to send RST/ICMP error messages back,
  and only do it then.
- Another performance improvement in the networking code by skipping
  ipsec tag checks if there are no ipsec flows.
- Large performance improvement for pf(4) by keeping information
  previously stored in a mbuf tag in the mbuf header directly.

Maurice

[Johan Beisser]

On Jan 15, 2008, at 9:34 AM, Stuart Henderson wrote:

>> I thought the performance improvement came from 4.1 with the  
>> removal of per
>> packet interrupts.
>
> http://www.openbsd.org/42.html
> Huge performance improvements in the network stack, including:
> # In pf, store routing table ID, queue ID etc directly in the packet
> header mbuf instead of using mbuf tags (which use malloc'd memory).  
> This
> yields a 100% improvement in pf performance.
> # Packet forwarding can skip IPSEC stack if no IPSEC flows are  
> defined.
> This yields a further 5% improvement in packet forwarding performance.
> # Skip TCP/UDP/ICMP/ICMP6 checksumming when not necessary. This  
> yields a
> further 10% improvement in pf performance.

Hmm. I'll do a test upgrade later this week, and once again try to  
knock my latency up to something kind of insane.

>> The closest relevant information from plus42.html:
>> * Enable interrupt holdoff on DP83816 sis(4) chips. Significantly  
>> improves
>> performance of such devices under load.
>
> that doesn't help your 83815D.

I know this.