[wuestenbrot]

zz, on Jul 11 2004, 03:48 AM, said:

Another limit could be if the user has uploaded more than downloaded during the session.

Cant understand, how this statement is connected with PS
Could someone help, please?

[zz]

If that limiter is to prevent people from creditshaping by setting files to powersharing, then that session limit would be proper (allow ps if uploaded more than downloaded).

If you just want to be able to automatically turn off powersharing when you consider the file to be well spread, then a check of spreading/complete sources would be better. But that should in that case be an option, not enforced.

/zz

[Motte]

zz, on Jul 11 2004, 10:50 PM, said:

If that limiter is to prevent people from creditshaping by setting files to powersharing, then that session limit would be proper (allow ps if uploaded more than downloaded).

I don't see how, or i should ask : so you want to forbid PS to non 'strict releaser' users ?

Quote

If you just want to be able to automatically turn off powersharing when you consider the file to be well spread, then a check of spreading/complete sources would be better. But that should in that case be an option, not enforced.

If no enforced max value, forget the idea. It would then be totally useless, and will still encourage credit shaping.

Sorry, but even if i love the PS functionnality, i really hope official devs will never include your PS in official client because it's really too powerfull. Lot of not well aware users put all their files in higher priorities ; hopefully, they mainly use the credit system, so the CS disturb the priorities. This weakness here, is beneficial. But if they could put all their files in PS ... they'll do it ... and ...

[stormit]

Hi MotteDeTerre

How are you my friend, you have some valid points as always. I myself only power share completed files thats all and you are right powershare is very powerfull. I hope that zz would improve this great feature that I love it really does the job. and ofcource one for zz .

[wuestenbrot]

Well, just have to repeat:
Powershare only available on complete files would prevent abuse.

[Andu]

It makes sense for incomplete files as well in certain situations but only if the amount of complete sources is really low.

[Motte]

Andu is right.
As far as i'm concerned i have no need of PS for uncomplete files, but in some cases it can be usefull. And i agree the max limit for uncomplete files has to be lower (and low) than the one for complete files.

Your welcome Stormit

This post has been edited by MotteDeTerre: 12 July 2004 - 12:29 PM

[Andu]

Maybe two different ratios for different behaviour. 1:3 if you do things properly and 1:2 on incomplete files with many full sources.

[MadlyMad]

hmmmmm

well, as others... i only use powershare for complete files, and 90% of times for 5 subtitles files =)

i just used once powershare for a incomplete file, situation :
near 100 sources (including me) waiting for last 2 chunk incomplete, since 2 months ! on a 45-50MB live video file
and only one complete source... online one day per week (and less ) AND really bad shared (big queues rating... :/ ), i was one of the first guy to receive one of the last 2 chunk, so i put it on powershare for 2 reasons :
1) to make the download faster for others
2) to prevent complete source to upload again the same chunk just before disconnecting... and so to prevent to re-waiting for a week (at minimum)... :/

and it worked great

well now, imho, i dont think the hard/static limit that you are talking about has many advantages...
well, sure it is another security if unaware people (hihi thanks motte ... didn't thought of this word... better using this than "newbies and assholes" lol ) put powershare on a file very long time
OR just to prevent human mistakes =)

okey
but, i m sure this limit is useless if the way of limiting powershare (that is i talked before) is well done


here is the thing :

most (not to say All of them !) of the european and american files on edonkey are shared like this :
for a file which have 1000 Total sources, most of times there is between 333 and 500 completes sources... and sometimes only 250

well, for asian files it is different, most of time 1000 Total sources files have 750 completes sources !!!!!

thats why I talked about a ratio 1:5 , between complete sources and total sources for a same file
maybe more (like 1 complete source for 6 total sources for example(when i mean "total sources", i mean "complete + incompletes sources" ))

really maximum would be 1:10 imho
1:5 seems really good to me ! under this (1:4; 1:3 etc...) would means this limit has no sense, and powershare could be almost always use !!! (or just "Always" use !)


my thought is to put a security to automatically disable powershare on a file, in certain situations.... so that official emule team would finally agree to add it in eMule!!
But with no way to have this feature entirely automatically (like auto-put a file in powershare mode because it is rare or something...) there already is auto-priorities for this !.... auto-powershare could be really too dangerous !

[leexgx]

its in morph mod and i think its on by default

[Andu]

Yea but I think MadlyMax has a point. I'm not sure if 1:5 is good enough. I think it should be 1:10. You still have release priority after that. Powershare is there to pump the file into the network. If you have a 1:10 ratio between seeds and leeches you have achieved what powershare is needed for. And if you insist on pumping the file you can just as well live with a 1:2 session ration. I know that releasers wouldn't mind anyway.

Btw I would probably get rid of rule 1 or put it to a low value like 10 or something like that. It's only important for rare files really and a rare file uploaded to 10 sources already makes it well distributed.

This post has been edited by Andu: 13 July 2004 - 12:34 AM

[MadlyMad]

MadlyMad

and i totally agree, 1:10 could be find to, cause of the "release priority" after...
but it seems to be a little too hard for me
it means, powershare will stop for example when there is only 100 complete sources for 1000 total sources
seems really hard to me... the file need to be powershare/spread again
i think

and, for the first rule, 25 seems good to me, a file with 10 complete sources is still a rare file... thoses sources can unshare quickly so...
for little files too (still thinking of subtitles for example)
many guys download them and delete them once they have been "use" without thinking it can be bad for the life of the file on the donkey network
so 25 seems good to me =)

and, with a first rule value of 10... and a 1:10 ratio for the second rule
it means powershare will auto-disable when there is 10 completed sources for 100 total sources !!! which is a really bad spreading for a file, and it can be again dangerous for his life on the network...

[Andu]

You are talking as if the full sources were the only clients with chunks with a 100:1000 relation you usually have every chunk spread so often across the network that you have a very high count of virtual full sources.

Same works with 10:100 especially if you have hideOS and SCS running. The distribution of chunks should be so well that you can leave it up to the network itself to distribute the file further.

[MadlyMad]

yes, it's true
i'm a scared boy

but, you are right IF the file is well shared
which is really not sure
10 ten sources is really low and many of them can disconnect.... unshare the file or just share it really bad (which can be almost as bad as unsharing it)

I released a few files, and I saw guys who unshare the file Just after completed it !!! .... while the file was in the very very beginning of the spreading/releasing period !!!
and not just one guy

really, I personnaly wouldn't feel safe if powershare auto-disabled itself when the file i'm releasing reach 10 completed sources for 90 uncompleted ones.



PS : HideOS ??? SCS ?

This post has been edited by MadlyMad: 13 July 2004 - 07:29 AM

[Motte]

MadlyMad, on Jul 13 2004, 09:25 AM, said:

PS : HideOS ??? SCS ?

A lot of mod use it. Have a look in SF-IOM subforum for feature explanations.

[CiccioBastardo]

I really do not like these limitations.

I use Powershare to share small files (smaleer than 1 MB), and I like the idea of those clients getting that file as soon as possible instead having to wait days (as it happens to me when I got them).
I also love when I can get 100KB files in 30 secs since I have added it to my download queue.

That's beyond the number of people sharing the file. Rare files have usually all complete sources, but often these sources keep the file in very low priority (voluntarly or not). This means you could get some good chance to get the file in resonable time, but in reality you keep on wating because they can't really share.
If just one of them would powersahre such files the number of sources would be pointless (1 would just suffice). That's what I call good sharing.

Yes, I agree only complete files should be powershared (and allowing automatic powersharing for files under a certain dimension, like for example 1/2 chunk).

[MadlyMad]

true

there is 2 kind of rare files :
- one with a "lot" (or not) of sources and really bad shared (and not really spread)
- one with no many sources but almost all are complete sources

in second way, (especially if the file is old) it is put on low priority
And ! .... the fact that almost all sources are complete ones, makes the download even more long... cause of no credits sharing possibility
hope to be clear

I m not sure a unlimited rule would be good to exists for a certain category of file
but, IF so, i would say a limit between 50 and 100 KB MAXIMUM ! 3MB is damn too high imho

but i still prefer to see a sort of limit as we talked before... for Every kind of file... Every size etc etc...

actually, sure it is important to download as fast as possible a file (especially for very small size i agree), but really most important is to keep the spirit of the donkey network compare to others networks.... it's to keep files alive As long as possible !!

[Galmok]

I have a question about the code in CUploadQueue::RightClientIsBetter or perhaps it is rather a question about the Visual Studio C++ compiler:

bool CUploadQueue::RightClientIsBetter(CUpDownClient* leftClient, uint32 leftScore, CUpDownClient* rightClient, uint32 rightScore) {
    if(
       (leftClient != NULL &&
        (
         (rightClient->IsFriend() && rightClient->GetFriendSlot()) == true && (leftClient->IsFriend() && leftClient->GetFriendSlot()) == false || // rightClient has friend slot, but leftClient has not, so rightClient is better
         (leftClient->IsFriend() && leftClient->GetFriendSlot()) == (rightClient->IsFriend() && rightClient->GetFriendSlot()) && // both or none have friend slot, let file prio and score decide
         (leftClient->GetPowerShared() == false && rightClient->GetPowerShared() == true || // rightClient wants powershare file, but leftClient not, so rightClient is better
          leftClient->GetPowerShared() == true && rightClient->GetPowerShared() == true && // they both want powershare file
          (
           leftClient->GetFilePrioAsNumber() < rightClient->GetFilePrioAsNumber() || // and rightClient wants higher prio file, so rightClient is better
           leftClient->GetFilePrioAsNumber() ==  rightClient->GetFilePrioAsNumber() && rightScore > leftScore // same prio file, but rightClient has better score, so rightClient is better
          ) ||  
          leftClient->GetPowerShared() == false && rightClient->GetPowerShared() == false && //neither want powershare file
          rightScore > leftScore  // but rightClient has better score, so rightClient is better
         )
        ) ||
        leftClient == NULL // there's no old client to compare with, so rightClient is better (than null)
       ) &&
       (!rightClient->IsBanned()) && // don't allow banned client to be best
       IsDownloading(rightClient) == false // don't allow downloading clients to be best
      ) {
        return true;
      } else {
        return false;
      }
}

My question is regarding the parsing of || and &&, more precisely, which of the two take precedence. In ANSI C || and && are equal and if the terms arent wrapped in () then the order of calculating is entirely up to the compiler. I can find a sequence such as this:

leftClient->GetPowerShared() == false && rightClient->GetPowerShared() == true || leftClient->GetPowerShared() == true && rightClient->GetPowerShared() == true &&

As we really don't know the order these terms are and'ed/or'ed together, the output is really undefined.

My point is really that we should always put () around groups belonging together and never trust "term1 && term2 || term3" to be evaluated like that. It could be evaluated as "(term1 && term2) || term3" but also as "term1 && (term2 || term3)" which makes it a completely different piece of code. It is entirely up to the compiler to evaluate and it doesn't necessarily do left-to-right or right-to-left evaluation unless it is told to.

Because of this, I would really like to ask you (zz or anyone knowing the Visual Studio .NET 2003 compiler) to put () around each pair of terms (or more if they are seperated by the same type of boolean operator). I may be able to do it correctly but I prefer a bit of help.

[MoNKi]

Look at this page on msdn, && has precedence.

http://msdn.microsof..._.operators.asp

This post has been edited by MoNKi: 17 August 2004 - 08:43 PM

[Galmok]

Oh, thank you for the information. Nice that that particular problem has been fixed. :-)

I will try to read the code in that light and hope it makes better sense to me now. ;-)