[Top][All Lists]

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

Re: [GNUnet-developers] Cadet bug: blocked cadet channel in case of non

From: t3sserakt
Subject: Re: [GNUnet-developers] Cadet bug: blocked cadet channel in case of non reliablle channel
Date: Sun, 24 Feb 2019 22:07:38 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.4.0

To be honest I did not checked if the oldest or the newest message is removed. I only looked onto the comment stating "Drop the oldest message in the buffer"

On 24.02.19 22:02, Schanzenbach, Martin wrote:
As far as I can see, the head element of a DLL is removed.
Unless elements are inserted at the tail (not the default) then the most recently cached message is evicted from the buffer.

Can you point me to the code you think this is happening?

On 24. Feb 2019, at 21:56, Christian Grothoff <address@hidden> wrote:

Signed PGP part
I agree with both of you, (1) default is in-order, so we may skip
messages, but not break the order. But if t3sserakt is right that the
code drops a more recent message in favor of an older message, that is
also terrible and should not happen.

That said, I do remember that that entire unreliable messaging was never
properly tested...

On 2/24/19 9:50 PM, Schanzenbach, Martin wrote:

a quick look into the bug (not a CADET expert) makes me questions the proposed behaviour:

"Proposal how to change that behavior:

We will not drop the oldest message in the queue, but we send as much messages from the queue as we have messages with consecutive MIDs. After that the queue is empty, or we again wait for the message that is missing now. Means we have to set the expected MID to that MID, because we are now waiting for another message to arrive."

Now, looking at the API of CADET, this channel has the following description:

  * Default options: unreliable, default buffering, not out of order.

Ergo, messages are _not_ delivered out of order. But that seems to be what you propose?
The transport is unreliable. So if you need any other behaviour, don't you just want a different OPTION? There are a few to choose from with that behaviour, no?


On 24. Feb 2019, at 21:33, t3sserakt <address@hidden> wrote:

Signed PGP part
Hey *,

please have a look onto this finding:

If nobody has a veto, I would change the behavior of non reliable cadet
channels, as I proposed in the bug description.



GNUnet-developers mailing list

GNUnet-developers mailing list

Attachment: signature.asc
Description: OpenPGP digital signature

reply via email to

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