RevConnect 0.401a Released

Know of something that might be useful to the DC community? Post it here! (Still, no advertising)

Moderator: Moderators

Locked
liny
Posts: 30
Joined: 2003-11-01 09:18

RevConnect 0.401a Released

Post by liny » 2004-04-19 19:16


GargoyleMT
DC++ Contributor
Posts: 3212
Joined: 2003-01-07 21:46
Location: .pa.us

Post by GargoyleMT » 2004-04-19 19:42

Is multisource safe yet?

liny
Posts: 30
Joined: 2003-11-01 09:18

Post by liny » 2004-04-19 21:07

GargoyleMT wrote:Is multisource safe yet?


for expert like you, it's safe.
for newbie or idiot it's not safe. because no $getmeta yet.
why not dc++ do $getmeat first, but ADC?

Twink
Posts: 436
Joined: 2003-03-31 23:31
Location: New Zealand

Post by Twink » 2004-04-20 00:10

liny wrote:
GargoyleMT wrote:Is multisource safe yet?


for expert like you, it's safe.
for newbie or idiot it's not safe. because no $getmeta yet.
why not dc++ do $getmeat first, but ADC?


so bascially you're saying that hashing isn't integrated into multisource stuff yet?

liny
Posts: 30
Joined: 2003-11-01 09:18

Post by liny » 2004-04-20 00:54

you are wrong.
RevConnect has changed from ed2k to tth.
But tth is only used for search, can not be used to verify file...
Because no way to get peer's hash tree unless dc++ god do it for us.

Qbert
Posts: 73
Joined: 2003-06-07 03:12

Post by Qbert » 2004-04-20 01:08

liny wrote:you are wrong.
He was only asking; notice the "?"

So I assume from this information then that RevConnect is just like it was a while ago, but at least working now, and also now with an added feature of searching by TTH?

GargoyleMT wrote:Is multisource safe yet?
liny wrote:for expert like you, it's safe.

So again nothing has been changed from its original implementation. So therefore its still not safe? (Don't get me wrong; I'm not saying its horrible or unusable if this is true.)
My Visual Studio .NET 2003 is licensed under my name, and the same for my operating system... What about you?
I surf on an OC3 without limitations, two to be exact, and I'm not joking.

ivulfusbar
Posts: 506
Joined: 2003-01-03 07:33

Post by ivulfusbar » 2004-04-20 01:17

I will, it is horrific and ugly and broken in most possible ways.
Everyone is supposed to download from the hubs, - I don´t know why, but I never do anymore.

liny
Posts: 30
Joined: 2003-11-01 09:18

Post by liny » 2004-04-20 02:59

I don't want to hurt anyone.
Maybe I misused english. Please forgive me.
If you have interest, you can try my client.
If you dislike, please comment it gently and justly.

GargoyleMT
DC++ Contributor
Posts: 3212
Joined: 2003-01-07 21:46
Location: .pa.us

Post by GargoyleMT » 2004-04-20 10:34

liny wrote:But tth is only used for search, can not be used to verify file...
Because no way to get peer's hash tree unless dc++ god do it for us.

True, but you can at least use TTHes, returned by search results and in XML file lists, to be an excellent guide to file compatibiliy. If the roots are the same, incremental verification isn't necessary.

Anyhow, I'm mostly referring to this experiment by OrangeSlice:
http://dcplusplus.sourceforge.net/forum ... php?t=7486

He found that the Reverse Connect code in ZDC corrupted most of his downloads, and as such, we've been discouraging users from trying any product based on your code since then.

Even before hashing, a rollback system was a pretty "safe" hack to ensure that downloads weren't corrupted.

Is RC now using $GetZBlock to get segments?

liny wrote:If you dislike, please comment it gently and justly.

If Reverse Connect respects file integrity, there's no reason t hate it. However, if your code will corrupt transfers though picking incompatible sources (or otherwise), your client is bad for the DC network. Multi-source has to be safe before any responsible coder would make it public.

liny
Posts: 30
Joined: 2003-11-01 09:18

Post by liny » 2004-04-20 20:01

GargoyleMT wrote:True, but you can at least use TTHes, returned by search results and in XML file lists, to be an excellent guide to file compatibiliy. If the roots are the same, incremental verification isn't necessary.


Yes, tth is helpful to pick out wrong sources.
If you only use sources with the same hash, the possibility of corruption will be very very low.

GargoyleMT wrote:Anyhow, I'm mostly referring to this experiment by OrangeSlice:
http://dcplusplus.sourceforge.net/forum ... php?t=7486

He found that the Reverse Connect code in ZDC corrupted most of his downloads, and as such, we've been discouraging users from trying any product based on your code since then.


The reason of curruption is bad/wrong sources. NOT the code.
I recommend you try yourself and let more users try and make comments.

GargoyleMT wrote:Even before hashing, a rollback system was a pretty "safe" hack to ensure that downloads weren't corrupted.


I don't agree this. Rollback is useful, but it is not a pretty way.
Many users still got currupted download by using DC++.
By see the tth of search results many users have the same hash of currupted download.
The reason of corruption which RevConnect can't avoid, current DC++ can't avoid too.
If I make a fake share, I promise rollback system will not work anymore, if you download from me.
DC++ with 'rollback' is better than nothing, but not much.

GargoyleMT wrote:Is RC now using $GetZBlock to get segments?


No.
1) my users are still using old version of DC++
2) my current algorithm doesn't fit $GetZBlock, more effort is needed

GargoyleMT wrote:If Reverse Connect respects file integrity, there's no reason t hate it.


No, what a lot of people hate and fear is multi-sources.

GargoyleMT wrote:However, if your code will corrupt transfers though picking incompatible sources (or otherwise),
your client is bad for the DC network.
Multi-source has to be safe before any responsible coder would make it public.


If 'rollback' is pretty, you should not worry about that at all. :)
All I need is something like BCDC++'s $GetMeta but implemented by DC++
It's not responsible to save tons of leaf hash in disk and do not use them.

GargoyleMT
DC++ Contributor
Posts: 3212
Joined: 2003-01-07 21:46
Location: .pa.us

Post by GargoyleMT » 2004-04-20 20:25

liny wrote:The reason of curruption is bad/wrong sources. NOT the code.
I recommend you try yourself and let more users try and make comments.

If the source blindly copies sources together - I.E. doesn't ensure that at least the boundaries overlap and are the same, then the problem is the source (for not being robust with bad input). If all the corruption is occuring in the middle of a segment, I agree, that there was nothing you can do about it.

liny wrote:By see the tth of search results many users have the same hash of currupted download.
The reason of corruption which RevConnect can't avoid, current DC++ can't avoid too.
If I make a fake share, I promise rollback system will not work anymore, if you download from me.
DC++ with 'rollback' is better than nothing, but not much

Rollback is a nifty hack. But it solves the problem, superficially, without having to extend the system.

It's nice to see the corruption inherent in the system, finally. I hope that we can bring this to user's attention and users will care about file integrity now...

liny wrote:1) my users are still using old version of DC++

I understand that. Isn't it more about what version of clients that others use, though? $GetTestZBlock has been enabled since 0.302. :)
liny wrote:No, what a lot of people hate and fear is multi-sources.

Agreed. New converts want it, and old users think it will subvert their precious slot system. Hubops fear that min-slot rules will cease to have meaning. (I'm sure I've missed some other points too.)

liny wrote:It's not responsible to save tons of leaf hash in disk and do not use them.

It is a disapointment that 0.307 didn't flesh out the hashing system. What's done is done. As new hash-related feature trickle in, maybe the users who dislike them will see their usefulness.

liny
Posts: 30
Joined: 2003-11-01 09:18

Post by liny » 2004-04-20 21:40

Thanks for all of your helpful comments and suggestion from the first day I came here. :)

GargoyleMT
DC++ Contributor
Posts: 3212
Joined: 2003-01-07 21:46
Location: .pa.us

Post by GargoyleMT » 2004-04-22 18:15

liny wrote:Thanks for all of your helpful comments and suggestion from the first day I came here. :)

:) Thanks for taking my complaints in a kindly manner. It's nothing personal, the kademlia thing is nifty, as is your implementation of eMule's secure ID system. I just wish that the multi-source was 100% never going to introduce corrupted files back into DC. Many, many, many users want this and will use your mod - it sucks if they're all helping to make DC more kazaa like through corruption.

Locked