Re: [Mldonkey-users] Investigation: No download for some, full downloads for the other

[Pierre Etchemaite]

Le Tue, 24 Dec 2002 01:56:21 +0100, Martin <address@hidden> a écrit :

(Sorry if I'm replying to that post, I couldn't find the post it's referring
to...)

> MLdonkey wrote...
> >       )  (if c.client_source_for = [] then !current_files 
> >           else c.client_source_for);
> > 
> > and tell me if your downloads are improved...

(Is that the "major bugfix" of the latest CVS commit ?)

Well, that doesn't look too good:

* it looks hackish, "if we don't know what this source is good for, ask for
everything", unless you can explain cases were it is the only thing mldonkey
can do;

* it will flood Lugdunum servers' test connections just as well as the plain
List.iter (...) !current-files

* here's how I think unknown incoming peers are greeted, tell me if my
expectations are wrong:
  - when they connect, query_files is called; since the peer is not
    known to be the source of anything, file requests aren't sent, and the
    client_last_filereqs timer is not updated
  - hopefully the incoming peer, at least, knows why it connected for, so it
    sends its filerequests
  - that calls add_new_location, that adds the file requested to the
    client_source_for, and calls query_files again(*), so the first
    filerequest, at least, is immediately echoed. Is it fine to send
    filerequests at that time ?


(*) oops, that was removed from my code ;)