Re: [lwip-users] lwip 1.4.1 send a sporadic RST ACK

[Joel Cunningham]

> On Oct 25, 2017, at 22:19, Axel Lin <address@hidden> wrote:
> 
> 2017-10-26 4:03 GMT+08:00 address@hidden <address@hidden>:
>> Joel Cunningham wrote:
>>> 
>>> 
>>> I've updated the remaining comments to mention the same priority on git
>>> master
>>> 
>> 
>> Hmm, thinking about this, shouldn't we instead change the killing scheme to
>> not kill active connections? It's been like it is for long, but given this
>> bug report, I think the way the comment says it would be might be a better
>> choice...
> 
> I use netconn APIs.
> I found it always create tcp pcb with TCP_PRIO_NORMAL.
> i.e. the priority is all the same.
> 
> So in such case, killing the active connections with the same priority
> or not does matter.
> In current code, create new connection won't fail because it will
> always kill the oldest active connections.
> Such behavior seems strange because I would expect new connection to
> fail rather than killing existing active connections.
> 

It does seem that killing the same priority as the new connection is harder for 
the app programmer to correctly account for because it means they have to limit 
their new connections before exhausting the pcbs.

Given that this is not an expected behavior of a traditional socket layer, i.e. 
a call to socket() will abort an existing connection, it may be even less 
likely that the programmer would account for this behavior 

Joel