[JDEV] File transfers

Marco Stolpe fireglyph at gmx.net
Fri Jun 7 18:17:20 CDT 2002


Ups, sorry for not using my real name in my last posting.

Well, now, after I've read Michael F Lin's posting, I've recognized 
myself that - when looking at the physical properties of some devices 
- my rules make a lot of assumptions which might not hold true.

One has to take into account many other criteria for making a good 
decision with regard to efficiency (network use and the time the user 
has to wait for a file). I've been much too theoretical, sorry.

Anyhow, to generalize a bit, my point was that maybe one could make 
any use of the information clients can exchange with each other 
before the transferal, including general properties like 
speed/available bandwith and data depending on time like the actual 
load, to make better decisions with regard to the downloading 
"strategy" to choose. (Instead of always going for the server 
solution!)

Some filesharing applications are already using information about the 
devices involved (56K modem, T1, T2, and so on). So maybe the 
question if we should use the server or not could be decided case by 
case, extending the if-then-else rules I've already written, in the 
style "if A=56K and B=T2 and load(A) > load(B) and load(B) < 
load(Server)" - or something like that. (The rule probably doesn't 
make sense, of course. ;-))

It might be interesting to adjust such rules either manually or 
automatically based on tests with real networks, using several 
solutions and configurations with different loads ... hmm ...

Only some thoughts,
Marco



More information about the JDev mailing list