On Thursday 24 January 2008 00:14, Robert Hailey wrote:
>
> On Jan 23, 2008, at 5:52 PM, Matthew Toseland wrote:
>
> > On Wednesday 23 January 2008 17:43, robert at freenetproject.org wrote:
> >> Author: robert
> >> Date: 2008-01-23 17:43:11 +0000 (Wed, 23 Jan 2008)
> >> New Revision: 17211
> >>
> >> Modified:
> >> trunk/freenet/src/freenet/io/xfer/BlockTransmitter.java
> >> Log:
> >> actually send the all sent notification
> >
> > IIRC we don't actually use this atm?
>
> In any event, now it's more debugged. But actually, the
> BlockReceiver's packet loss notification is written around getting
> this message. Now... in theory the link layer is reliable, so this
> should not be used (any packets re-ordered would eventually be
> received). And indeed, in my testing of this most of the reordered
> packets are eventually received, however, I would also get very-
> occasional hangups ("terminating send, we haven't heard from the
> receiver in +2m"), which (absent some connectivity issue) would
> indicate that the receiver is still waiting for more packets.
>
> So yes... allSent should be redundant with the presence of allReceived
> and reliable delivery, but as best as I can tell something is still
> afowl... and this appears to have fixed it (or worked-around it in the
> case of a link layer bug).
Hmmm. Please debug the link layer bug! If we are losing messages then that's
*bad*, it will cause timeouts etc. Really we should be able to use a simpler
data transfer layer (e.g. Bulk*) because we have a reliable transmission
layer. Let me know how you get on, if you need help / aren't interested in
the bug then file a bug with as much as you've got so far and what tools
you've been using (simulator e.g.).
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20080124/58d0a7fa/attachment.pgp>
