Re: Fwd: Re: IP: Microsoft breaks Mime specification

On Tue, 22 Jan 2002 22:39:06 PST, Ed Gerck said:

> When there is more market value in breaking a technology than following it,
When your corporate strategy is "We know what you want, and we will do
illegal and questionable things to ram it down your throat and make you think
that's what you wanted", it's a moot point. 

Or maybe it's simply that you as a vendor have been made painfully aware
that you've managed to bork the BGP specs - but actually *fixing* the problem
isn't a 3-line change because it requires re-architecting a large part of
the code base, and your management is telling you to devote your efforts
to writing the code to support the feature that Sales has already sold
to customers(*)....

> what is your answer?  I find it hard to believe that Microsoft implemented
> that broken MIME because its engineers can' t read RFCs or did not bother
I don't find it hard to believe at all.  In this case, I think it was
a *bug*, pure and simple.  Some code-monkey didn't consider that a certain
set of headers could appear together, and their testing probably didn't
find it because Outlook doesn't create that combination of headers.

I won't discuss some of the comments I got when I was trying to get
the Exmh MUA to generate Content-Disposition: headers for the
various parts of a multipart/signed so if the remote MUA didn't understand
the /signed and degraded it to a /mixed (as per spec), that it wouldn't
do something ugly with the application/pgp-signature (which would almost
certainly be degraded to application/octet-stream).  Yes, I can read the
RFC's.  That doesn't mean that I got the code "right", or that the authors
of the remote MTA had the same reading of the RFCs as I did...

Microsoft's variant implementation of Kerberos however...

> to.  It seems to me that we do need to put non-conformance on the spotlight.
Yes, non-conformance needs to be put in the spotlight.  However, I don't
think we need a formal structure to it..

> When a flaw is qualified as a flaw and made public, it may embarass its
> creators enough to make them change it.  Left alone, the flaw will probably
> become a feature.
I've seen enough of the various trade journals run news articles of the
form "X doesn't play nice with Y".  Amazingly enough, quite often these
are followed by articles of the form "Vendor of X rushes to ship fix".

The system works.  The only thing the IETF as a group needs to do is to
make sure Microsoft *does* eventually issue a fix for bug OfficeQFE:4781,
and that shouldn't be *too* hard to do now that it's acknowledged to be a bug.
-- 
                                Valdis Kletnieks
                                Computer Systems Senior Engineer
                                Virginia Tech

(*) No, I *dont* know how Cisco managed to break the BGP spec, or why/how
it isn't fixed in the field.  But I'll admit *I* have shipped production
code that still had "This is wrong but hopefully will work anyhow" comments
in it.  Yeah, the code is wrong.  Yeah, it mostly works anyhow.  Guess
what the chances are that it will get fixed correctly? ;)

pgpH1jZUeU5UW.pgp
Description: PGP signature