|
It seems to me that there is a problem with the current
scheme for marking fragment. This is what the existing draft says:
Fragmentation (F):
When set(1) the attribute is fragmented. When clear (0), the
attribute is not fragmented. When the value portion of an extended
attribute exceeds 246 octets the value is fragmented over one or more
extended attribute. In this case the Fragmentation Flag is set on
the attribute(s) that contain the value fragments. It appears to me that this means that all of the
fragments are marked as such. If so, suppose that 2 extended attributes
of the same extended type are adjacent in a packet & both are
fragmented. How can you tell where the first ends & the second
starts? I can think of 2 easy ways to fix this: either specify that the
flag is cleared in the attribute containing the first fragment or in the
attribute containing the last fragment. Note that the latter option
basically turns the bit into a "more" flag, which has some precedent
in IETF protocols ;-). Any opinions/other suggestions? |