Re: [Nmh-workers] Wishlist item - beefing up slocal pattern matching

[Mike O'Dell]

OOPS..... now you've stepped in it...

Having been through all this before, the first time 25 years ago
when RFC-733  was revised to give RFC-822, and all the
machinations since then...

both the specs and operational reality treat headers as
case-independent.  *if* you wish to honor case in header
processing involving pattern-matching, it MUST NOT be the default.
"slocal" in particular must honor arbitrary CaPiTaLiZaTiOn.

(0) the header tag to the left of the : is ALWAYS case-independent.
        the spec specifies how you *should* send it, but it
        required that the recipient to accept the header in spite
        of bizarre capitalization.

(1) anyone who created a new header with a case-sensitive data portion
        (to the right of the :) made a very, very bad design decision

(2) yes, i know base-64 can be used to encode the value of some
        extension headers (eg, X-foo), but that merely
        requires not screwing with the data in the header,
        which is already a requirement. if the processing
        of extention-header semantics wishes to honor case,
        that's fine, but *nothing else* is obliged to do so.

(3) ergo, it's still a bad decision to make headers case-sensitive

The fact that Postini made a very poor decision is no recommendation.

NOTE: Personally, I have always been on the side of honoring case in headers.
Especially now that we've admitted there are symbols outside the
26 letters of the US Roman alphabet, "case-folding" is really stupid
because it only applies to 26 symbols out of the 32,000 one can
use in Unicode-clean software.  I also happen to have an apostrophe
in my last name - a fact that almost always betrays the brain-damage
in Microsoft IIS web sites - and know many people with much more
exotic diacritical marks in their name, so i'm quite sensitive to
the "6-Bit-DEC10-ASCII-ization" imposed in the email RFCs.
(It wasn't that long ago that lower case was operationally
verboten in email text.)

however, the specs are the specs, and Postel's Robustness Principle:

        "Be conservative about what you send and
         liberal about what you receive."

is still very, very good advice.

        cheers,
        -mo

Igor Sobrado wrote:
> In message <address@hidden>, Joel Reicher writes:
> > I remember a while ago that people were wanting the slocal pattern
> > matching to be full regular expressions. I still think that is a good idea,
> > but it's occurred to me that the pattern matching of a maildelivery file
> > and the search criteria of pick do much the same job, so I'm now thinking
> > it would be a good idea for pick and slocal to share that machinery.
> > Any enhancement to pick would be an enhancement to slocal.
> >
> > Since pick can do full boolean expressions, I think a new maildelivery
> > syntax might have to be introduced. That's the only part of this idea I
> > don't like.
> >
> > Thoughts?
>
> I would like pattern matching to be case sensitive.  There are some
> antispam services (e.g., Postini) that rely on capital letters on
> user defined fields in the email header (those starting with "X-").
> An example: http://spam.acm.org/public/filters/faq/postini_transf1.html
>
> Don't know if standard fields in the message headers can be capitalized
> in different ways, but it will certainly be useful to allow case-matching
> at least as an optional feature.  On the other hand, Unix users are
> accustomed to use the right capitalization for commands, filenames,
> and so on...  ;-)
>
> Just a wish...
>
> Igor.
>
>
> _______________________________________________
> Nmh-workers mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/nmh-workers