Foil 7 ("What we are not doing..."): One could say that when the DTF picks a protocol from the lists supplied by source/sink
it is doing a primitive form of protocol negotiation. I do not see it as part of bandwidth negotiation.
the "or" in this bullet is wrong. It should either be "and" or the two parts should be separate bullets (better)
Foil 8 (DEPR). I would replace "YOU" with something like "DTI instance".
Mario's comment about security has opened an interesting question: is there a security issue in passing DEPRs around if
they contain authentication information? I can see one but it may be the same as my ability to disclose information
that I have access to to a bad guy. This needs further thought.
Foil 9 ("Where do DEPRS come from"). See my note from last week. I do not think that DMI should specify where they come from
but I do think that we have an obligation to suggest what data services need to supply. This may be something
that OGSA Data needs to add to its architecture document.
Foil 10 ("DTF"). I would be careful about "best" in the first bullet. I would rather say that the DTF just picks one. "Best" has far too
many dimensions and they are out of scope for DMI.
"Negotiaste" -> "negotiate".
I agree with Mario that the future items may end up outside of DMI. But I think that DMI is the logical place where
these issues should first be considered even if DMI decides to push them to some other part of the
architecture.
Foil 11 ("DTI") I agree with Mario - we should include the state diagram.
Allen Luniewski
IBM Cross Brand Services
IBM Silicon Valley Laboratory
555 Bailey Ave.
San Jose, CA 95141
408-463-2255
408-930-1844 (mobile)