[Gzz] Some thoughts about Gzz p2p layers

[hemppah]

Okay, as said earlier, my focus in thesis is on locating the data and
downloading it (as Tuomas said, *obtaining* the data). As I have done more
research, I see three logical sub-processes when obtaining a data from
distributed p2p network  (some existing examples at the end of each line):

1) WHICH: identify which data you want to get basd on meta-datas etc. (e.g. FASD
in Freenet). In our case, it's block IDs and urn-5s, right ?
2) WHERE: find out where the data is (system does this) (e.g. Chord, CAN,
Kademlia and other routing systems, if based on some unique ID)
3) HOW: how I should download it (e.g. BitTorrent, MFTP)

Of course, in some existing systems, these sub-processes are merged, e.g.
Freenet merges the WHERE and HOW, and uses FASD to WHAT. Gnutella merges WHAT
and WHERE, and uses HTTP to HOW.

Here are some questions:

1) Should we think this whole process as described above ?

2) Is it wise to have seperate logical layers to obtain the data from network
(modularity vs. "more work") ?

3) It seems to be, that for DHTs, the WHERE (and WHAT) layer is ok (mature
enough). For me, the question is, how we should implement WHAT, or more
important, is this a problem why we should not use DHTs ? However, for instance
in Overnet (based on Kademlia), keyword searches are *possible* (and not very
hard to implement, as discussed with Benja, based our assumption). 

4) How much we can affect by HOW to WHERE: e.g. hotspots in DHTs (MFTP in
overnet --> less hotspots or no hotspots at all ??).

For clarification, is my focus more on WHAT and WHERE, instead of WHAT ?



-Hermanni



-------------------------------------------------
This mail sent through IMP: http://horde.org/imp/