[Top] [All Lists]

Re: Impending changes to "body" draft.

2004-11-10 14:48:25

On Wed, Nov 03, 2004 at 11:45:40AM -0500, Cyrus Daboo wrote:
If any document authors will not be at the meeting, either in person or via
jabber, could they please send me privately a summary of the state of their
document so I can present that as part of our discussions, thanks.

After the IETF, I plan to repost the "body" extension draft in
preparation to send it towards workgroup last call.

Two open issues:

- :binary.  I added it in the last iteration, and now
  yanked it out again.  It's a neat hack, but I feel
  like I can't make this work as a serious feature,
  especially in connection with variables and
  wilcard matches.

  The hex space-separated hex pair values look ok to me,
  graphically, but are unique in the E-mail world.  Cyrus
  suggested making them =-separated to allow reuse of
  quoted-printable engines, but the format I was describing
  is not _quite_ the same, I find =-separated hard to
  read, and I think people will assume that where
  quoted-printable works, there's a way of doing
  base64-encoding, and now we're really getting ridiculous.

  I don't like the ability my format adds to match against
  single nybbles.

  The main use of this - to match against virus signatures -
  really is better served in virus scanners.

  QUESTION: Are we okay with throwing out :binary, or
  is someone using it for something worthwhile?
  If we want to keep it, how strongly do we feel about
  using a format that looks more like quoted-printable?

I say "lose it".

- In the unpublished version on disk now (and appended
  to this message), I've added an explicit exemption from
  the variable-setting side effect of matches.  That makes
  body much easier to implement, but I hate special
  cases and this is one.   We need to agree that we do
  think the implementation cost of wildcard-matches with
  variable-assignments in body is so high that we don't
  want to do it.

  QUESTION: Is it okay to have body :matches and
  :regex scans not set variables?

I think this is a very good idea.


<Prev in Thread] Current Thread [Next in Thread>