[intser]

Hello.

I work in qMule development team. It is new cross platform eDonkey/torrent client. It based on libtorrent_rasterbar library and libed2k library.

libed2k features:
1. Cross platform
2. High speed
3. Async IO operation.

In plans:
1. IPV6 support
2. Multi servers

[Zangune]

intser, on 17 January 2013 - 11:31 AM, said:

In plans:
1. IPV6 support

I hope people will work together or will use a compatible implementation

DavidXanatos, on 10 January 2013 - 03:41 PM, said:

I'm looking forward to adding IPv6 support to my ed2k/eMule capable File Sharing client

---

intser, on 17 January 2013 - 11:31 AM, said:

2. Multi servers

eMule developers idea about this

Unknown1, on 20 July 2003 - 05:01 PM, said:

Features you won't see in eMule:

• Multiple Server connections - We already talk to all the servers. Creating extra connections is just wasteful and loads the servers.
• Server hopping - We already talk to all the servers. This is just wasteful and loads the servers.

It's really necessary?

This post has been edited by Zangune: 17 January 2013 - 11:44 AM

[DavidXanatos]

intser, on 17 January 2013 - 10:31 AM, said:

Hello.

I work in qMule development team. It is new cross platform eDonkey/torrent client. It based on libtorrent_rasterbar library and libed2k library.

libed2k features:
1. Cross platform
2. High speed
3. Async IO operation.

In plans:
1. IPV6 support
2. Multi servers

About IPv6 I'm designing also a protocol extension for that, some first thoughts can be found at http://forum.emule-p...howtopic=155981

It would be great if we could agree on a standard so that the implementations will be compatible.

About servers you should drop server support and instead go for KAD only,
as it happens I can help you with that a lot: http://update.neoloa...ces/MuleKad.rar this is a MuleKad process, it contains kad only with minimal changes it can be compiled on any C++ platform as all main parts use only clean C++ with STL.
You can use it as separated process accessible via Pipes or sockets if your plroject is not GPL or you can copy and past the KadCode Core into your project if it is GPL, this way you wouldnt need process separation.

The MuleKad module bases on the most recent aMule code so its state of the art.

How about Adding Nat- Traversal, on my todo list its quite high up, a compatible implementation here would also be great.

David X.

[hooligan3000]

some bad mods had multi server..
but its useless all ed2k server have the same information

[YumeYao]

I love the way you seperate ed2k as a single module(I hope libed2k doesn't take extra dependencies), while I personally dislike qt because of its ugly(non-native) look and bloated overheads.

But... why is libkademlia missing? Currently kad is fairly more important than ed2k.

Also I believes the offcial eMule has a great space for performance/speed improving, and so do its mods. For aMule, I haven't checked its code too much but just from the face, it seems aMule has made some effort improving code quality.


IPv6 is great but it should have a common design of protocol extension supported by any kinds of ed2k/kad clients. I still have no thoughts on IPv6 and haven't experienced any coding on it.


@DavidXanatos:
404 error for MuleKad.rar.

[DavidXanatos]

YumeYao, on 17 January 2013 - 03:07 PM, said:

@DavidXanatos:
404 error for MuleKad.rar.

Thats strange for me it works, can you go on; http://update.neoloader.to/ ?

[YumeYao]

DavidXanatos, on 17 January 2013 - 11:14 PM, said:

Thats strange for me it works, can you go on; http://update.neoloader.to/ ?

something like a server default page.

[DavidXanatos]

you seam to have something messed up with oyur DLS provider, try: http://www.opendns.com/

[xilolee]

I have the same YumeYao's errors.
Using my gateway as dns and using Google DNSs, Windows firewall disabled (Yes, I'm using it).

This post has been edited by xilolee: 17 January 2013 - 04:14 PM

[YumeYao]

DavidXanatos, on 17 January 2013 - 11:54 PM, said:

you seam to have something messed up with oyur DLS provider, try: http://www.opendns.com/

Did you mean DNS?

I use google's DNS server (8.8.8.8) as my primary dns server. I've just tried opendns:

C:\Users\YumeYao>nslookup update.neoloader.to 208.67.222.222
服务器:  resolver1.opendns.com
Address:  208.67.222.222

非权威应答:
名称:    update.neoloader.to
Address:  94.23.146.14


C:\Users\YumeYao>nslookup update.neoloader.to 208.67.220.220
服务器:  resolver2.opendns.com
Address:  208.67.220.220

非权威应答:
名称:    update.neoloader.to
Address:  94.23.146.14


C:\Users\YumeYao>nslookup update.neoloader.to 8.8.8.8
服务器:  google-public-dns-a.google.com
Address:  8.8.8.8

非权威应答:
名称:    update.neoloader.to
Address:  94.23.146.14

[DavidXanatos]

Thats odd, may be my domain provider fucked something up,
on my PC it works but it may have some old dns results cached.

Here you have a rapidshare download to the relevant files:
MuleKad (GPL): http://rapidshare.co...573/MuleKad.rar
and the required NeoHelper Library (LGPL): http://rapidshare.co...5/NeoHelper.rar

[YumeYao]

Rapidshare says:

Download permission denied by uploader.

[DavidXanatos]

ok now it should work try again

[YumeYao]

Cool. just found a google cache copy of your site: http://webcache.goog...5&hl=en&ct=clnk

I like the idea to seperate each module very much, it seems that MuleKad is based on aMule's kad module.

For the fact aMule (and this module) uses std::string as its string type, I don't know whether qt has its native string type as CString in MFC or wxString in wxWidgets. It's said that std::string is not as efficient compared to other string types listed above.
However, if I were to rewrite this part, I may choose to completely avoid using a C++ string type. C++ strings are too heavy and they only benefits when the string is to be modified a lot(I think it's not the case for kad). Also, if another part of the app uses a different C++ string type (say, std::string and wxString), the conversion also takes couples of memory copies...

Well, I understand performance is not the top topic for eMule. But it's just my style. lol.

[DavidXanatos]

Qt actually has its own string type QString. Qt is a full standalone framework it does not use any other frameworks like wxwidgets or boos or whatever.

The MuleKad implementation used std::string/std::wstring there is a interface class that handles Kad and performs the conversion form std strings to QString.

The nice thing about the MuleKad module is that aside of the said interface class it is kept entirely framework free it uses only stl so it can be easily integrated in any other project without the need of adding any additional framework.


For MuleKad the separation into modules makes only sense if your project is not GPL.
In case of qMule as far as I understand it the ed2klib is BSD and the GUI app itself GPL, meaning that the muleKad module couldnt be included directly in the ed2klib but there could be an interface so that it is linked through the GPL's gui app. A separated process would only be needed for users who want to compile the lib into an non GPL application.

The reason other kad modules are separated like neo kad is that this modules are designed to be at least in principle usable by multiple applications at once.
So The Neokad Module for example can be used for NeoLoader for filesahring stuff, and form som other app for something else, like P2P DNS or whatever, ist general purpose.


David X.

This post has been edited by DavidXanatos: 17 January 2013 - 08:10 PM

[YumeYao]

I see.

I know QT is a full standalone fw I mentioned wxWidgets just for comparison.

Yes, almost every C++ string type other than std::(w)string has conversions from and to std::(w)string, but in this case memory copy is performed and unnecessary overheads may take place.

Say you have:

void foo(std::string &str1);

QString bar()
{
  QString qstr1;
  ...
  // serveral processes on qstr1.
  ...

#if 1 //You may write:
  foo(qstr1);
#else //The compiler may produces:
  std::string str1 = qstr1;
  foo(str1)
  qstr1 = str1;
#endif

  ...
  // serveral processes on qstr1.
  ...
  return qstr1
}

So for performance, a single app uses only one C++ string type may win some points.

As far as I see, kademlia could be re-written without any C++ string types(std::string, CString, wxString, QString, etc.).

A C++ string type benifits following (compared to C pointer):

• Comparing: It could be compared using '==', '>', '<', etc. , therefore could be used in maps(Consider maps is a template and if you use pointers... strings may never get compared properly)
  
• Processing: It has dynamic allocating and management, therefore makes string processing much human-friendly.
  
• Return Type: When used as a return type, its allocating and freeing are handled transparently, reduces the risk of memory leaking.

But as for kademlia module, it benifits none of above, that why I would vote for using C string pointers in kademlia.

[DavidXanatos]

>But as for kademlia module, it benifits none of above, that why I would vote for using C string pointers in kademlia.

Well you could write a kademlia implementation that does not use c++ strings thats true (actually the Torrent DHT lob neo uses has been writen in C only). But as far as I can see you couldn't port the current MuleKad code to a string free version with any reasonable amount of afford, I think you would be faster writing a new implementation from scratch.

but here the topic is qMule, and as far as I see thay are using boost librarys and thats kind a STL on steroids, so eider it uses stl strings anyway or you could port is using only find and replace.

David X.

[intser]

Zangune, on 17 January 2013 - 11:41 AM, said:

intser, on 17 January 2013 - 11:31 AM, said:

In plans:
1. IPV6 support

I hope people will work together or will use a compatible implementation

DavidXanatos, on 10 January 2013 - 03:41 PM, said:

I'm looking forward to adding IPv6 support to my ed2k/eMule capable File Sharing client

---

intser, on 17 January 2013 - 11:31 AM, said:

2. Multi servers

eMule developers idea about this

Unknown1, on 20 July 2003 - 05:01 PM, said:

Features you won't see in eMule:

• Multiple Server connections - We already talk to all the servers. Creating extra connections is just wasteful and loads the servers.
• Server hopping - We already talk to all the servers. This is just wasteful and loads the servers.

It's really necessary?

I think it will be in libed2k as feature. GUI completely separated software and programmer can use this feature or not.

[intser]

YumeYao, on 17 January 2013 - 03:07 PM, said:

I love the way you seperate ed2k as a single module(I hope libed2k doesn't take extra dependencies), while I personally dislike qt because of its ugly(non-native) look and bloated overheads.

But... why is libkademlia missing? Currently kad is fairly more important than ed2k.

Also I believes the offcial eMule has a great space for performance/speed improving, and so do its mods. For aMule, I haven't checked its code too much but just from the face, it seems aMule has made some effort improving code quality.


IPv6 is great but it should have a common design of protocol extension supported by any kinds of ed2k/kad clients. I still have no thoughts on IPv6 and haven't experienced any coding on it.


@DavidXanatos:
404 error for MuleKad.rar.

libed2k it is completely separated crossplatform library for eDonkey protocol. It dependents only on boost. Kademlia is simple different library. GUI was developed on Qt, because based on qbittorrent client.

[Neoo26]

With open DNS all neoloader page are working, only the forum is not working (error 404).