mldonkey-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Re[2]: [Mldonkey-users] Some features and debianization


From: Goswin Brederlow
Subject: Re: Re[2]: [Mldonkey-users] Some features and debianization
Date: 06 Oct 2002 01:12:57 +0200
User-agent: Gnus/5.0808 (Gnus v5.8.8) XEmacs/21.4 (Honest Recruiter)

Rainer Krauss <address@hidden> writes:

> Hi Goswin,
> 
> Saturday, October 5, 2002, Goswin Brederlow wrote:
> 
> GB> The querying stuff should be non optional since that just wastes the
> GB> servers (and for you more importantly) your bandwith.
> 
> GB> Downloading a file only after you've seen it once is a protection
> GB> against broken files. Quiet a lot of files are incomplete in the
> GB> donkey network and will most liekly never complete even in month.
> GB> Implementing a "force" option is not realy high on my todo list since
> GB> I see no point in it. Downloading an incomplete only spreads it around
> GB> further without making it more complete.
> 
> hmmm... i doubt that this is a good idea.
> 
> I see 2 main drawbacks. First, what about new Files? Think of a Movie
> with 3 CDs which often occures. If everyone waits until seen completed
> it would spread very very slow. To get your download started you must
> at least one time connect to the original uploader until others have
> it full. Sounds like a snowball effect to me or, in this case, no
> snowball effect anymore.

Thats not how it works. When a client tells mldonkey that it has some
chunks mldonkey notes the current time for those chunks. A few chunks
here, a few chunks there. If that array of times does not contain any
0.0 (initial value) anymore every chunk has been seen somewehre at
some time.

Then you can start to download.

It does not wait for a client that has all chunks, that would realy
delay things.

> The second point are rare files. They will most likely be very much
> harmed from spreading as you must find the complete file before you
> take place in the chain of shares. So you cant start a partial
> download and share what you get to others. If there is only one
> uploader with the full file you must wait until you find him to be
> online and then the download starts - this will just let you have a
> long wait and a lot of people giving up. And chances are good that you
> cant finish your download because the uploader gets offline before.
> Chances would have been better if you allready downloaded the parts
> which are available and download only those not beeing available so
> far until now from him. Again kind of missing snowball effect will
> make rare files vanish fast.

The point is that you wait till you have seen at least one source for
every chunk. For a very rare file you must connect to the uploader
once, thats true. But then you would download one of the rare
chunks. If the downloader goes offline then you keep downloading more
chunks from other places and you probably just got the one chunk
everyone is missing (you searched for the file to be seen complete for
some time, so you know whats chunks are rare [does not quite work that
way but true enough]).

Initialy there is a delay before the downloads starts and its gonna be
big for sparse files. For normal files its a few seconds delay, for
sparse files maybe a few days. But such sparse files will download
slowly because the person(s) having the missing chunks are offline all
the time. An extra day delay in a week weigthed against knowing that
its not a broken file that will never finish sound reasonable.

Out of my currently 500 files there are only 2 still waiting to be
seen completly and they are <2days old.

MfG
        Goswin




reply via email to

[Prev in Thread] Current Thread [Next in Thread]