Re: [tcpm] poll for adoption of draft-ananth-persist-02

["Anantha Ramaiah (ananth)" <[e-mail]>]

Hi Mark, 

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On 
> Behalf Of Mark Allman
> Sent: Thursday, April 01, 2010 12:53 PM
> To: Murali Bashyam
> Cc: [e-mail]
> Subject: Re: [tcpm] poll for adoption of draft-ananth-persist-02
> 
> > There are other aspects too, it's also important for the 
> application 
> > to control the duration that the connection spends in 
> persist state, 
> > this is planned to be accomplished via a socket option.
> 
> Why?
> 
> I mean, sure, an app may want to not let a connection hang 
> around forever.  But, can't it just ABORT whenever it wants?

Yes, the application can ABORT the connection whenever it wants. 

The socket option is suggested for multiple reasons (use cases) :-

- orphaned connections, where the application has already closed the
connection and the connection is in persist state.
- Typically when the data cannot be sent (due to retransmisions) there
is a retransmit timeout happening and a error ETIMEDOUT returned to the
application, for persist case there is no way for the application to
know about the connection not making any progress. When this socket
option is set, TCP would behave the same way as would for ETIMEDOUT
case.

The reason why the socket option is been suggested because we want to
preserve the default behaviour of persist forever but an override
suggested by the application if needed by means of a socket option. (or
sysctl or something equivalent). This ensures that we are not violating
RFC 1122.

Also this is an informational draft and typically one would want to
explain the situation better and some implementation advice, that is the
reason we have put the socket option in the appendix. Hope this
explanation conveys the functionality of the socket option, do you think
some clarfication are needed in the document?. (actually we have
received other comments too and will put all of them together when we
submit the next revision).

-Anantha

> 
> allman
> 
> 
> 
>