On Sat, 26 Apr 1997 10:13:57 -0700 Dave Crocker
<[log in to unmask]> wrote:
> The bulk of Mime
> environments dispatch to the content-aware application based solely on the
> content-type/subtype string. That is, they do not use any additional
> parametric information to make the dispatching decision.
>...
> Otherwise, there will need to be a
> generic sub-switch that is called by the Mime processor and then has the
> job of looking at the parameter, to determine the actual application to
> invoke. This latter approach is, of course, highly suboptimal.
Dave,
While I think your comment about "the bulk of MIME
environmnts" is certainly correct, I think that two
different issues arise here. On the second, the choice
needs to be made by subject matter experts.
(i) It seems to me that we've either got to get
applications to start looking, early on, at parameters, or
we need to either dramatically constrain the discussion of
what parameters are for (or remove them entirely) before
MIME goes to full standard. That would be a drastic step:
since I think parameters are useful, we need to push a bit
on the implementation approaches.
Of course, whether
examining parameters is suboptimal or not depends on the
assumptions and quality of the implementation. Forcing
parameter processing onto a system that normally discards
or ignores parameters would, without question, be painful.
Handling them in an implementation that processes them
normally involves trivial marginal work (This would seem to
me to be an opportunity for an IMC interoperability test.)
(ii) For MARC records, I suspect the situation is much
like some of what we've found for EDI. The important thing
may be to distinguish things that are MARC records from
things that aren't, then to pass the former off to
processors --processors that will rarely be conventional
email engines-- that know what to do about them. It is
possible that such processors might reject some subtypes as
uninteresting or incomprehensible, but they are probably
the right level at which to make those decisions.
Anyway, Ray and others, Dave's comments and my reply above
pretty well identify the tradeoff. You need to look at
likely usage and configuration patterns and figure out what
would work best. I suspect that one of the key
discriminators might be to ask whether, if a new MARC
format came along that you wanted to process, it would be
supported by upgrading (or installing modules into) a
"MARC" application or whether the interpretation machinery
would more likely to be something that you would want to
hang off the email system (or some MIME-processing web CGI,
or something similar).
john
|