how timeout recommendations from rfc2821 apply when pipelining rfc2920

06/11/2006, 21h30

Say for example a smtp client batches "mail from" and 4 "rcpt to:" and
one "data" to a smtp server.

If the smtp client is using RFC timeouts, would that mean that the
client should wait up to 27 minutes for a response from the server,
since the server could be pipelining the response back, as per rfc2920?

Should the smtp server break apart a pipeline message on timeout
boundaries? After all, it has no way of knowing what the client
timeouts are.

If it assumed RFC timeouts, then it could notice that the response to
the mail from and the first couple rcpts are ready, but the others are
not and the timeout is getting close, so flush the response back and
continue proccessing the rest.

09/11/2006, 13h41

jmaimon@ttec.com wrote:
> Say for example a smtp client batches "mail from" and 4 "rcpt to:" and
> one "data" to a smtp server.
>
> If the smtp client is using RFC timeouts, would that mean that the
> client should wait up to 27 minutes for a response from the server,
> since the server could be pipelining the response back, as per rfc2920?
>
> Should the smtp server break apart a pipeline message on timeout
> boundaries? After all, it has no way of knowing what the client
> timeouts are.
>
> If it assumed RFC timeouts, then it could notice that the response to
> the mail from and the first couple rcpts are ready, but the others are
> not and the timeout is getting close, so flush the response back and
> continue proccessing the rest.

http://www.jmaimon.com/sendmail/patc...v1.81308.patch
http://www.jmaimon.com/sendmail/patc...sh-response.v1

This alters sendmail's behavior when a client uses PIPELINE.

Any thoughts?

06/11/2006, 21h30	#1
jmaimon@ttec.com Messages: n/a Hébergeur:	how timeout recommendations from rfc2821 apply when pipelining rfc2920 Say for example a smtp client batches "mail from" and 4 "rcpt to:" and one "data" to a smtp server. If the smtp client is using RFC timeouts, would that mean that the client should wait up to 27 minutes for a response from the server, since the server could be pipelining the response back, as per rfc2920? Should the smtp server break apart a pipeline message on timeout boundaries? After all, it has no way of knowing what the client timeouts are. If it assumed RFC timeouts, then it could notice that the response to the mail from and the first couple rcpts are ready, but the others are not and the timeout is getting close, so flush the response back and continue proccessing the rest.

09/11/2006, 13h41	#2
jmaimon@ttec.com Messages: n/a Hébergeur:	Re: how timeout recommendations from rfc2821 apply when pipelining rfc2920 jmaimon@ttec.com wrote: > Say for example a smtp client batches "mail from" and 4 "rcpt to:" and > one "data" to a smtp server. > > If the smtp client is using RFC timeouts, would that mean that the > client should wait up to 27 minutes for a response from the server, > since the server could be pipelining the response back, as per rfc2920? > > Should the smtp server break apart a pipeline message on timeout > boundaries? After all, it has no way of knowing what the client > timeouts are. > > If it assumed RFC timeouts, then it could notice that the response to > the mail from and the first couple rcpts are ready, but the others are > not and the timeout is getting close, so flush the response back and > continue proccessing the rest. http://www.jmaimon.com/sendmail/patc...v1.81308.patch http://www.jmaimon.com/sendmail/patc...sh-response.v1 This alters sendmail's behavior when a client uses PIPELINE. Any thoughts?

Outils de la discussion
Afficher une version imprimable Envoyer un lien vers cette page par email