Autor: W B Hacker Data: Para: exim users Assunto: Re: [exim] RCPT delays and PIPELINING
Jakob Hirsch wrote: > Quoting Philip Hazel:
>
>> But I am really interested in views as to whether doing the automatic
>> flush is a good idea or not.
>
> I am also in favour of the flush-before-delay. The client will get the
> responses for the previous commands, so he will have at least that, and
> the server will not appear to be completely stuck. The behaviour is more
> "natural", I'd say, and it's better for debugging.
>
> We could introduce modifiers for the delay keyword (like "delay =
> 10s/noflush"), if somebody wants the old behaviour, but it's probably
> not worth it.
>
Given the (potentially) torturous nature of a pipelined 'session' will it not:
- take a lot more coding and testing in Exim to insure nothing breaks than the
gain may justify?
- have a reasonable chance of finding 'appropriate' handling at the far-end?
The second part concerns me more than the first as:
- it is beyond 'our' control.
- many of the 'majority' MTA have rather poor reputations for doing things
'correctly' in general.
IOW - even given the tools, will we find a frequent need to make exemptions - or
even turn them off?
Personally, I'd rather give up delays if need be.... which breaks nothing.
