PATCH: longish HELO in dccm.c

Leandro Santi
Fri Oct 18 19:16:29 UTC 2002

Vernon Schryver wrote:

> > From: Leandro Santi <>
> > ...
> > The following patch fixes a problem with long smtp HELO arguments
> > in dccm.c. See the patch for more deailts, it is quite obvious in fact.
> > ...
> Thanks.  I'll include it in version 1.1.16.
> What SMTP client has a HELO value longer than MAXHOSTNAMELEN+30?

I'm not sure yet. I just know it happened because I've been running my dccpost
client for a week here with an assertion in that particular code path. Today the
assertion was triggered so I actually fixed the code.

> By chance, are you not stripping the ESMTP chatter from the HELO value?
> Or do you not use that big a buffer for HELO values?

No, I'm stripping it, just as milter does (I think). And the buffer size is just
equal to dccm's bufsize.

> The main change for 1.1.16 will be to let clients retry DCC reports
> as queries to other DCC servers.  The current tactic is just give up
> after a tries to a server that doesn't respond, to avoid reporting
> the same checksums more than once when the problem is not that the
> server is down but that the client can't hear the server's answer.
> This change seems interesting because there is a lot of mob behavior
> in the 1000 clients using the 7 servers answering
> In one hour, one of those servers will be receiving 5 requests per
> second while the others are less busy.  In the next hour, the busy
> server will be almost idle and the mob will be huddled around another
> server.  I theorize that is caused by transient high packet loss rates

Yes, I see some similar behaviour here, I think. Anyway, I'm very happy
with the DCC so far :-) The dcclib code has proven to be quite generic,
to the fact that I didn't touch a *single* line under the dcclib tree.


> in the Internet.  Such losses beg doing something to mitigate their
> other effects.
> Vernon Schryver

More information about the DCC mailing list

Contact by mail or use the form.