ietf-822
[Top] [All Lists]

Re: Content-Disposition Header

1993-06-29 10:03:48

On Mon, 28 Jun 1993 15:58:55 -0400, Keith Moore 
<moore(_at_)cs(_dot_)utk(_dot_)edu> said:


  moore> Which we could require, if we're willing to state that
  moore> there's never any circularities.  Can an object have both
  moore> "owner" & "requires"?  Can these things nest or be mutually
  moore> referential?  How hard it will be to implement depends on
  moore> this.  A general-purpose implementation has to be fairly
  moore> complex, but if we're willing to say that there's no cycles
  moore> (possibly because you can't have both owner & requires, or
  moore> possibly just because you aren't allowed to have cycles, so
  moore> cycles can be treated as errors) then a linear implementation
  moore> is more plausible, but the general mechanism is less
  moore> powerful.

This is a problem with any scheme that permits recursive (singly or
multiply) links; one must be careful not to be trapped in a loop.

  moore> I just realized that I might want to mail someone a set of
  moore> application/html docs that have cyclic references among one
  moore> another.  Hmmm...

This is not so bad with hypertext documents; html links have sort of
implicit 'attachment' semantics, in that the links are not resolved
unless the user requests it.

Consider instead a compound image format, in which images A and B
referred to each other. It would be like facing two mirrors at each
other, but a lot more resource-intensive... :)

Perhaps MIME parsers should use recursion counting, like namei(), to
avoid infinite loops?

-Rens



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