Add New Protocols

[sirinath]

Hi,

Pl. add protocol support for: FT, Ares, OpenFT, Turtle F2F, OpenNAP, Soulseek, Freenet, DC, MP2P, Metalink etc.

S

[old_death]

Well, no problem.

This will be done during the next month*.




*Condition: You help us getting the developers needed to do the code work. Above statement is only true if we are successful in getting these devs.

[ailurophobe]

Metalink was discussed before and it is possible since Shareaza already supports more of its features than most applications using it. All it takes is a developer who wants to code it.

Ares is possible, but the available source for it is written with Delphi, so converting from that to something Shareaza compatible would be a major pain. But if a developer willing to do the (lots of) work comes along then I don't recall anyone having serious objections to Ares during the previous discussions about this.

DC is not really useful. This was discussed few forums ago and IIRC the consensus was that adding DC to Shareaza gives no benefit over running both Shareaza and a separate DC client. Then again same is more or less true of integrated IRC support...

FT has been rejected several times.

[mojo85]

ROFL, Old_Death wow believed it until I noticed the small print. In all seriousness, folks when asking for more protocols please ask yourself if the 3 Networks (G1, G2, eD2k), and 3 protocols (HTTP, FTP, and BT) are not enough? Do they not get you what you want? What could possibly be an added benefit of Fast-Track aside the obvious of more virus and poor file integrity (weak hashes)?

Lets get real, in my opinion as it always has been we could do with less network ... and add a DHT layer. Get rid of G1 and eD2k ... they are old networks not worth supporting. Add mainline DHT to support BT and use mainlineDHT to offload SHA1 hashes found in G2. If we want to take it safe, then time to create a suitable DHT for G2.

We have not focused on G2, ever since its inception. We need to refocus on this network freshen it up with new ideas as it is our bread and butter, all other networks are add-ons. Time to clean house, I'm sure Ryo has some G2 spec revisions or improvements he would like to add.

http://g2.trillinux.org/index.php?title=Main_Page

There is Kevo's proposal that hasn't even been discussed (ASE --> Active Source Exchange)

[ailurophobe]

The benefit of a P2P is not dependent on how old it is, it is dependent on how many peers it gives you access to. And last I looked both Gnutella and ED2K were bigger, and hence more useful, than G2. Although if you can find a G2 source it is usually a good one. I am currently testing an eMule mod and lacking those very few G2 sources really is noticeable. (So far I am still on track to go back to using Shareaza. eMule, even modded, sucks.) So obviously I'd agree it would be a good thing if we could improve the G2. And you are right, ryo does have ideas about that, so do I, and many other people around. Probably good ideas too. But that has been true for years and I haven't seen much happening. What G2 lacks is leadership. You talk about Gnutella and ED2K being old protocols and networks, but actually those networks have been making steady technical progress over the years. Maybe not as much as BT has, but the contrast to G2 is pretty deep.

Anyway, for G2 the first step should be getting more applications using it. Having a cross-platform hub only application would be a good start. DHT is not magic. BT and ED2K need it. Gnutella benefits from it. G2 already has a non-broken network model. It could get by with just better hubs. Specifically better QHTs. The current system is effectively capped at 20 bits and sucks at routing word queries. It can't tell the difference between having the search terms in the same file and ten files having one term each. This sucks if you have a large library. Or if you are a hub with 200+ leaves. Adding a bloom filter would help a lot. And AFAIK put G2 ahead of the kademlia based DHTs in word queries. I could go on and on, but the point is not whether my rantings are practical, the point is that there really is much that could be done to G2 and that it should start with the hubs and that the easiest way for that is to have a dedicated hub separate from Shareaza. And a lightweight client to test changes. So Shareaza would only need to add leaf support for any new features. (You could bundle the dedicated hub with Shareaza, though.)

[sirinath]

How about pluggable protocol support and some tools to implement the plugins. Then any body interested can implement what they might fancy and perhaps experiment using new experimental protocols altogether using Shareaza as a framework.

[old_death]

[octagram-3]

[ailurophobe]

Synchronizing the share is simple enough, you just tell Shareaza to share the correct folder. And people should be able to choose which networks they use anyway, no big difference linking to two installers or telling people to enable and configure an extra network. Also the available bandwidth to share the file does not depend on which networks are used, so the scaling does not depend on whether your downloaders use one or both networks. In theory splitting the networks causes source availability problems, but the kinds of situations where source availability is a serious issue do not really change. So either way you have to do the same things to improve it.

With search you have a better point. But isn't DC different enough that most people would have a preference on which network they want to use? And use different networks for different files?

I probably should explain that while something like having both DC and other networks in the same app does improve your download by increasing available sources, it does not increase your upload which means that it does not improve overall speed of the network and if not done properly can lead to networks becoming slower over all by making leechers faster. Which is something P2P developers care about. So if networks are different enough in design to make combining them non-trivial, it is better not to do it without a compelling reason. (Like the popularity of BT, for example.)

[octagram-3]

[ailurophobe]

Shareaza still supports non-NTFS drives and was IIRC originally written for Windows 9x support. So using ADS wasn't really practical. By the time it became practical Shareaza was already designed to do without. Shareazas watch folder ability really is quite good. I tried using eMule (which doesn't) for a while and it was a major pain.

Different networks -> different services -> different applications. I get what you mean though.

People have patterns in what kinds of files they look for. And which kinds of files they share. This results in different networks usually being much worse or much better depending which type of a file you are looking for. And one or another network being better bet to look for a specific file. But like I said with search you do have a point. Search aggregation has clear value. Personally I mostly use link databases for non-casual purposes and if the link is in the web, you can use Google.

First you list things DC would add and then you ask how it is different? Truthfully it has been years since I last checked out DC so I am not really qualified to discuss the technical details. Who knows how much has changed? Not me for certain. And not like I kept notes and accurately remember what the details were back then either.

AFAIK Shareaza has no way to prefer local/fast peers at the moment.

[octagram-3]

I mostly list features of clients, not the protocol itselves. I have impression that GreyLink and Shareaza are extending in the opposite directions (towards each other).

When I browse LimeWire hosts, I can see no structure. It looks like Shareaza introduced directory structure into Gnutella (or maybe Shareaza doesn't understand LimeWire filelist format?). However, this is early implementation of this feature and it's not that useful as I wish. This is of little use compared to GreyLink having almost got things right.

GreyLink introduced TTH comments and MediaInfo queries. However, MediaInfo queries has a very limited use. I can query info of a single file (codec, length, tags), but I can't select all the tracks of a specific artist (excluding the case when user keeps music in order) because DC doesn't track metadata. Whenever somebody sends a metadata request, metadata are being calculated from scratch and are not being saved anywhere. This is of little use compared to Shareaza organizers.

Once again, almost everything I list is features of clients and not protocols. There are indeed some technical issues. E. g. when you look for Eiffel 65 Follow Me, GreyLink will find Eiffel 65\Follow Me.mp3 , Shareaza won't (didn't checked, just guess). But can't Shareaza be changed a bit? GreyLink can also find folders "Eiffel 65" when I'm looking for just "Eiffel 65". I don't think this difference to be a stopper. Or any other difference.

[raspopov]

GreyLink is a DC++ client with very bad manners i.e leecher, attacker (!!!), fake tagger,fake sharing, slot blocker, has cryptographics, has no sources.

[kevogod]

[octagram-3]

[diztrancer]

[raspopov]

[profyler1]

I agree with all the replies.
There are no need for all of these networks, BUT i cant say that i would be sad if i see soulseek and Ares networkj in Shareaza.
For Ares, (I know, its on delphi language and so to complicate to implement in c++) because of its increase popularity. (almost 4 to 5 times of shareaza at least here in brazil. . And i know the results may be bad like for morpheus with riaa puting fake and currupted files on its servers.

And for Soulseek for having a lot of rare mp3s that are very hard to find on Shareaza with only few sources. i actually have the soulseek p2p only for that reason on my pc.

I liked a lot the implementation of DC++ and im using it constantly. I think i was a awsome job that u guys did, the best one maybe.

Like others, im always afraid of shareaza loosing to newer networks (so teoretically faster/smarter) beeing developed and only supporting the G2 as main network. I hope this never happens since im a shareazaholic.

I know this is not google neither microsoft. that everyone has a private life/job and shareaza being a community with limited developers/contributors, so i hope this will someday change so we se a hyper multi sharing plataform here!

And thanks for all the support i had and for making this awsome p2p program. I dont see much thanks to the ppl behind this project.