Date: Mon, 3 Aug 1998 09:57:32 +0100 (BST)
From: Philip Hazel <ph10@cus.cam.ac.uk>
To: Chris Newman <Chris.Newman@innosoft.com>
cc: Detailed Revision/Update of Message Standards <drums@cs.utk.edu>
Subject: Re: require complete syntax checks?
In-Reply-To: <Pine.SOL.3.95.980731095902.8530I-100000@elwood.innosoft.com>
Message-ID: <Pine.SOL.3.96.980803094620.1200B-100000@taurus.cus.cam.ac.uk>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Fri, 31 Jul 1998, Chris Newman wrote:

> Lax enforcement of syntax rules leads to "standards creep" where a server
> is sloppy and then poor clients take advantage of the sloppy server and
> stop working with stricter servers.  

Exactly. This is sadly the situation we are now in as a result of "be
generous in what you accept" in the past. I myself am guilty in this
regard. At one point I tightened up the syntax checking of the domain
name in HELO in Exim. As soon as that version was released I got many
complaints that it no longer accepted domain names containing
underscores, and I had to issue a patch. It was unacceptable to some 
organizations to have to modify dozens of clients that had previously 
worked find (and before Exim, had worked with Another MTA).

Once anything non-strict is at all widely deployed, it is very hard 
indeed to tighten up the rules.

> Then it becomes harder over time to
> write an interoperable implementation as there are all these sloppy things
> that never were in the standard (and most people don't want in the
> standard) that can help interoperability.

Quite! Maybe there is a case for somebody writing a program to test SMTP 
servers for this kind of thing, and for general conformance to the 
standards? 

> SMTP has seen far too much standards creep already, it would be helpful if
> SMTP implementors expended effort to collectively prevent more standards
> creep by strictly enforcing those syntax rules which haven't yet been
> violated in the field.

The only way to do that is to try to be strict, and then relax if 
necessary. It is a painful process because it causes annoyance.

> At this point, however, getting the revised specs done is far more
> important than rehashing this issue yet again.  So I personally oppose any
> changes to the current 821bis/822bis drafts with respect to syntax checks. 

I think I have to agree with that.


-- 
Philip Hazel            University of Cambridge Computing Service,
ph10@cus.cam.ac.uk      Cambridge, England. Phone: +44 1223 334714.