Re: [exim] Per-recipient post-DATA acknowledgements

Ian Eiloart wrote:
>
>
> --On 7 October 2008 19:49:44 +0800 W B Hacker <wbh@???> wrote:
>
>>
>> Am I the Lone Ranger in wanting to tilt toward an implementation ten
>> years in use vs one not yet even dry?
>
> NOPE. There are three suggestions so far: XEXDATA, XLMTP, and XPRDR
>
> I favour XPRDR technically (strict timeouts enable tarpitting without
> losing compliant hosts, for example),

Not sure that cannot be done 'easily enough' anyway... and w/o overly
relying on timeouts.

> but XEXDATA has the advantage of
> having an existing implementation.

The biggest advantage is that the Exim community has to do only half the
work - eg meet the same documented spec that courier already supports.

Finding a few courier-mta users who will test should be easy enough.

They'd like to get more use out of their feature as well.

Now - XPRDR would at first glance have the same advantage, and source
code is there for both of 'em.

But I just don't see that XPDR's 'parent' MTA has anything close to the
coverage or pool of admins of courier-mta which is in ports and packages
everywhere, and in very long-term production use, regardless of 'market
share'.

Once there are two players interoperable, the third one is easier to enlist.

Final and most enduring solution? There ain't no such animal.

> XLMTP has the advantage that LMTP is
> in widespread use, so there's lots of code out there.
>

ACK. - but no fewer than TWO MTA have to have an implementation created,
of the parts to be used in smtp. We can't just adopt lmtp directly as a
substitute.

AND they have to be independently debugged AND must interoperate.

The fact that it hasn't happened yet indicates it will not be easy to
make happen now.

> I agree with you that we should favour XEXDATA because it's already out
> there, even though I think it is inelegant.
>

Elegant or otherwise, XEXDATA is reputed to do what it was designed to
do in courier-courier interaction, and not just since last year, either.

> However, I'd also advocate supporting both EXDATA and XPRDR, because
> that might give us code that's flexible enough to be adapted to any
> similar scheme.
>

Can't fault that *so long as* it doesn't kill the whole deal.

As all the top F/OSS MTA are 'C' critters, it is all academic until a
'C' coder or three start chunking.

Bill