Message-Id: <v03101301af0d8d0a06e8@resnick1.isdn.uiuc.edu>
In-Reply-To: 
 <Pine.SOL.3.95.970122133401.24620K-100000@eleanor.innosoft.com>
References: <19970118052632.15868.qmail@koobera.math.uic.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Thu, 23 Jan 1997 16:02:55 -0600
To: Chris Newman <Chris.Newman@innosoft.com>
From: Pete Resnick <presnick@qualcomm.com>
Subject: Re: invisible continuation lines
Cc: Detailed Revision/Update of Message Standards <drums@cs.utk.edu>

On 1/22/97 at 3:48 PM -0600, Chris Newman wrote:

>On Sat, 18 Jan 1997, D. J. Bernstein wrote:
>>
>>> (1) RFC 822bis clients MAY interpret empty continuation line as a
>>> header/body separation.
>>
>>Objections:
>>
>> 2. The argument for your proposal is that it permits good handling of
>> BITNET-damaged messages. Philip Hazel has pointed out an algorithm
>> that achieves the same goal without corrupting 822-compliant messages.
>
>The algorithm is a heuristic, and thus doesn't always work.  For message
>display, treating the invisible continuation line as the end of headers
>guarantees no information loss (although potential for unnecessary
>uglyness). For filtering, doing what RFC 822 states is probably the best
>way to treat such lines.  For other purposes, Hazel's algorithm may be
>best.

Right. Two points:

1. In addition to the above, another reason not to use Hazel's alogorithm
is that it requires a two-pass parser, something that many people don't
have now in their products.

2. I don't think anyone prior to Chris has suggested using Hazel's
algorithm for header interpretation decisions (like filtering), but use the
"empty continuation = separator" method for display to the user decisions.
I think actually this turns out to be the "best of both worlds." Explaining
this point in the document might address everyone's objections: It would
make it clear what the real dangers are and would provide implementors with
the information they need to decide what to do.

What do people on both sides of the aisle think of this idea (in #2)?

pr

--
Pete Resnick <mailto:presnick@qualcomm.com>
QUALCOMM Incorporated
Work: (217)337-6377 / Fax: (217)337-1980