S M wrote:
Martin Stone Davis [EMAIL PROTECTED]
mailto:m0davis-yBeKhBN/[EMAIL PROTECTED]
wrote in message news:[EMAIL PROTECTED]
1. We probably should go with Toad's idea of dropping the HTL system
innb sp;favor of a timeout-based system. This is a radical change,
and needs to be
Martin Stone Davis wrote:
I just skimmed through http://www.ovmj.org/GNUnet/download/ebe.ps, which
describes GNUnet's economic model. In short, we would prioritize
requests based, in part, on how well requesting nodes have serviced our
requests in the past. The benefit would be a system in
A thought.. Why not give NGR development a couple of weeks break and try to
get muxing or something up and running in the mean time? When you have no
relly good ideas on some specific problem it can be good to move over to a
completely different matter for a while and then get back to tackling the
Same here.. I am running build 6366 and 40+ of the 50 estimators look
exactly the same.. However the look of that *the same* seems to be changing
pretty rapidly..
/N
- Original Message -
From: Ian Clarke [EMAIL PROTECTED]
To: Discussion of development issues [EMAIL PROTECTED]
Sent:
Niklas Bergh wrote:
A thought.. Why not give NGR development a couple of weeks break and try to
get muxing or something up and running in the mean time? When you have no
relly good ideas on some specific problem it can be good to move over to a
completely different matter for a while and then get
Toad wrote:
as load goes up, we make what we accept more and
more specialized. Overload pressure strengthens specialization.
Won't unobtanium routing make network analysis very easy?
If an attacker wants to determine the specialisation of a node, he will just
overload it with requests for random
The proposal for getting rid of HTL seems to be based on using time
expired rather than hops expired, but doesn't that require a globally
agreed time?
Ian.
___
Devl mailing list
[EMAIL PROTECTED]
Ok, the new stable build seems to be working quite well, are other
people experiencing the same thing?
We need to take stock of the situation with NGR. I think one problem
has been a willingness to dream up solutions, and implement them, before
actually understanding what the problem is.
I
No it does not. Read my proposial for Positive Trust baised Freenet.
The proposal for getting rid of HTL seems to be based on using time
expired rather than hops expired, but doesn't that require a globally
agreed time?
Ian.
___
Devl mailing
Postitive Trust == GNUnet modle.
My proposial impliments both.
S M wrote:
Martin Stone Davis [EMAIL PROTECTED]
mailto:m0davis-yBeKhBN/[EMAIL PROTECTED]
wrote in message news:[EMAIL PROTECTED]
1. We probably should go with Toad's idea of dropping the HTL system
innb sp;favor
Ok, the new stable build seems to be working quite well, are other
people experiencing the same thing?
quick comment: there is still some polishing that should be done to the current
stable codebase before 0.5.3 can go out. For example, 5046 wouldn't learn about
other new nodes through
Ok, I'm a bit behind in my e-mail so I just read this. But my postive trust
baised solution should solve this.
Hopefully this should shut up all those [EMAIL PROTECTED]* people, and you know who
you
are; who are talking about removing NGrouting, or saying that we should route
Nodes A,B,C,D all have 15 ecounds of trust in each other.
A wants some data.
Routs it to B with TTL of 10.
B docks A 14.14 trust.
1 secound passes.
B routes to C with a TTL of 9.
C docks B 12.73 trust.
1 secound passes.
C routes to D with TTL of 8.
D docks C 11.31 trust.
1 secound passes.
D
Ian Clarke wrote:
Ok, the new stable build seems to be working quite well, are other
people experiencing the same thing?
I don't think you can judge it yet. Only a fraction of users have updated.
I'm running a new (always) transient node on stable and it's routing
other requests. I thought a
Sorry, I should have been clearer.. I think that this 0.5.3 idea is a good
one but what I meant was:
1. Do the 0.5.3
2. Muxing or something
3. More NGR.
/N
- Original Message -
From: Ian Clarke [EMAIL PROTECTED]
To: Discussion of development issues [EMAIL PROTECTED]
Sent: Sunday,
OK, I should clarify. That was not directed at, you, Toad, Ian, Martin, Ed, Susa, or
any of the other contributers here. I acknowledge that NGrouting is complicated and is
hard to make work. I acknowledge there is a need for a working network even if it does
not work. I acknowledge the need for
Hi,
Am I correct in assuming that this model is fairly much independent of the
routing algorithm used, or am I missing the point completely? I understand
that with your proposal a node needs to have sufficient trust to be able to route
but I am not clear how you decide to route to a node.
Ed
Again, read my origional post. It explains all of these things and has code for how to
do much of it.
You get more trust for routing Faster so you route baised on speed and speed alone. So
you just use NGrouting but take out the DNF estimators. A DNF is intreprited as I
cannot get the data in
Thomas Leske wrote:
Toad wrote:
as load goes up, we make what we accept more and
more specialized. Overload pressure strengthens specialization.
Won't unobtanium routing make network analysis very easy?
If an attacker wants to determine the specialisation of a node, he will
just
overload it
[EMAIL PROTECTED] wrote:
Postitive Trust == GNUnet modle.
My proposial impliments both.
In which case, you'll have to answer Ian's objection:
And the disadvantage of such a system, as I have pointed out many
times in the past, is that new users get a terrible experience of
Freenet, which is a
[EMAIL PROTECTED] wrote:
Postitive Trust == GNUnet modle.
My proposial impliments both.
In which case, you'll have to answer Ian's objection:
And the disadvantage of such a system, as I have pointed out many
times in the past, is that new users get a terrible experience of
Freenet,
On Sun, 2003-11-30 at 08:55, Ian Clarke wrote:
Ok, the new stable build seems to be working quite well, are other
people experiencing the same thing?
We need to take stock of the situation with NGR. I think one problem
has been a willingness to dream up solutions, and implement them,
Ian Clarke wrote:
Ok, the new stable build seems to be working quite well, are other
people experiencing the same thing?
We need to take stock of the situation with NGR. I think one problem
has been a willingness to dream up solutions, and implement them, before
actually understanding what
Ian Clarke wrote:
snip
One thing that is important is simply to figure out how accurate NGR's
estimates actually are, and whether their estimates are statistically
significant.
If you look at the diff* diagnostics, I think you'll be convinced that
we are consistently wrong in one direction.
[EMAIL PROTECTED] wrote:
[EMAIL PROTECTED] wrote:
Postitive Trust == GNUnet modle. My proposial impliments both.
snip
Martin Stone Davis wrote:
You'll also have to answer the question I had about GNUnet-style
economic models: what crucial problem would be solved that wouldn't
be solved by the
I know that may not seem straight forward. But I didn't come up with this proposial in
an hour. I spent a long time thinking about the problems with the network, and decided
that time, rather than HTL was the way to go. as did you. That presented a lot of
problems too. (and security issues) The
Thomas Leske wrote:
Martin Stone Davis wrote:
The whole point of NGR is that nodes learn the specializations of
other nodes.
Yes, of course, other nodes should learn about the specialization of
other nodes. But they must not be able to do so without propagating
any data and spending a
[EMAIL PROTECTED] wrote:
I know that may not seem straight forward. But I didn't come up with
this proposial in an hour. I spent a long time thinking about the
problems with the network, and decided that time, rather than HTL was
the way to go. as did you.
Okay, fine.
That presented a lot of
Martin Stone Davis wrote:
I don't think this is a major problem. If the author realized this was
happening, the same document (essentially) could always be inserted at
multiple places in the keyspace.
A good target document for censoring would be the manifest of a DBR-freesite.
The key of the
Thomas Leske wrote:
Martin Stone Davis wrote:
I don't think this is a major problem. If the author realized this
was happening, the same document (essentially) could always be
inserted at multiple places in the keyspace.
A good target document for censoring would be the manifest of a
Also, I'd really like to test the black hole in stable as well; it should be
much less potent, but the never-QR-ing element is still there.
Sincerely, i think this would be pointless.
We already know that old stuff is subject to that attack. That anyone
can bring freenet down even with low
Aureliano Rama wrote:
Also, I'd really like to test the black hole in stable as well; it should be
much less potent, but the never-QR-ing element is still there.
Sincerely, i think this would be pointless.
We already know that old stuff is subject to that attack. That anyone
can bring freenet
We already know that old stuff is subject to that attack
No, we don't know if classic routing is. We suspect it might be to some degree, but we are not sure.
pgp0.pgp
Description: PGP signature
___
Devl mailing list
[EMAIL PROTECTED]
Zlatin Balevsky wrote:
We already know that old stuff is subject to that attack
No, we don't know if classic routing is. We suspect it might be to some
degree, but we are not sure.
Fine, but that's not the point. The point is to make Joe User happy
with stable so we don't kill off
On Sun, 30 Nov 2003 [EMAIL PROTECTED] wrote:
The idea is this: We eliminate the notion of HTL. Instead we replace
it with a trust based system with TTL.
...
void processRequest(Node node,Key key, TTL ttl)
{
float load=getCurrentLoad(); //Check our current load.
//Tenitively
On Sun, 30 Nov 2003 [EMAIL PROTECTED] wrote:
No it does not. Read my proposial for Positive Trust baised Freenet.
The proposal for getting rid of HTL seems to be based on using time
expired rather than hops expired, but doesn't that require a globally
agreed time?
If clocks are out of
On Sun, 2003-11-30 at 13:33, Thomas Leske wrote:
Martin Stone Davis wrote:
What bad thing would a malicious node operator do with that
knowledge?
He could censor a certain document. Assume he has the resources to
lauch a denail of service attack against a limited number of nodes. If he
Zlatin Balevsky wrote:
We already know that old stuff is subject to that attack
No, we don't know if classic routing is. We suspect it might be to some
degree, but we are not sure.
Fine, but that's not the point. The point is to make Joe User happy
with stable so we don't kill off
Martin Stone Davis wrote:
Okay, so the attacker could censor the current edition. But he wouldn't
be able to censor all of them, since they are distributed throughout the
keyspace. The reader could then just click to retrieve one of the
previous day's (or week's) editions. The fred interface
Ian Clarke wrote:
The proposal for getting rid of HTL seems to be based on using time
expired rather than hops expired, but doesn't that require a globally
agreed time?
Ian.
No, I believe the agreed time isn't global---it's agreed just between
two nodes at a time. I think the way Toad
Zlatin Balevsky wrote:
Zlatin Balevsky wrote:
We already know that old stuff is subject to that attack
No, we don't know if classic routing is. We suspect it might be to
some degree, but we are not sure.
Fine, but that's not the point. The point is to make Joe User happy
with stable so
On Sun, 2003-11-30 at 15:40, Thomas Leske wrote:
Martin Stone Davis wrote:
Okay, so the attacker could censor the current edition. But he wouldn't
be able to censor all of them, since they are distributed throughout the
keyspace. The reader could then just click to retrieve one of the
Ok, yesterday it didn't work - nodes didn't learn about each other, it
was busy, etc.
Today however, the request success ratio was more than 10% for a brand
new node (18 out of 124 externally requested keys successful), and frost
is chugging along (recorded a download speed of 100k at some
Fine, but that's not the point. The point is to make Joe User happy
with stable so we don't kill off donations. Experiments with black
holes in stable might make Joe mad.
No tv and no beer make Homer (ehm Joe) go crazy
No tv and no beer make Homer (ehm Joe) go crazy
No tv and no beer
I still think its a good idea to invest some toadtime into polishing it
before 0.5.3, but so far its kicking ass.
Yeah, just downloaded almost 300MB in half a day!
That's a new record for me with Freenet.
And we must consider that my stable node is completely new, I wiped
out everything I
Zlatin Balevsky wrote:
I still think its a good idea to invest some toadtime into polishing it
before 0.5.3, but so far its kicking ass.
That is good news, and yes - I suspect there are a few things that need
to be tidied up (for example, didn't someone mention that acquisition of
new refs was
I'm not sure the acquisition of new ref's is borked, I've watched my rt
shrink and regrow, so I'm pretty confident that that's not an issue, but
I have noticed that every ref in my rt eventually backs off, which isn't
very helpful to the cause, it also seems nodes aren't passing data
between each
The very idea of using a formula for making decisions about routing has
one major flaw and that is the innacuracy of the estimators. Unless a
perfect estimator is developed which will give the exact value of a
given variable, any formula will produce humonguous margin of error.
As you know,
Zlatin Balevsky wrote:
The very idea of using a formula for making decisions about routing has
one major flaw and that is the innacuracy of the estimators. Unless a
perfect estimator is developed which will give the exact value of a
given variable, any formula will produce humonguous margin
I have always thought that the whole NGR thing was misguided. Freenet works
because of keyspace specialization. NGR trashes keyspace specialization in
favor of speed, which reduces Freenet from an medium speed expressway to a
grid of city-streets with traffic-lights and congestion. You might be
Just looked at my nodes stats before I close it down for the time being
and I noticed this figure
Outbound connections that are to peers not in the routingtable
20.37037%
Why aren't these peers being added to the routing table? They obviously
are capable of dealing with request which
Thses
Histogram of requested keys.
This count has nothing to do with keys in your datastore
Dec 1, 2003 1:39:01 AM
keys: 478
scale factor: 0.4383561611175537 (This is used to keep lines 64 characters)
0 |===
1 |==
2 |
3 |==
4
52 matches
Mail list logo