- From: Miles Sabin <miles@milessabin.com>
- Date: Sun, 22 Dec 2002 12:05:28 +0000
- To: www-ws-arch@w3.org
Assaf Arkin wrote, > In fact, my experience, just like yours, proves that any multi-party > choreography can be broken down into a set of two-party interactions, > and any set of two-party interactions that is performed in the proper > order would result in a multi-party interaction. I think it's uncontroversial that from a technical point of view this transformation can always be done in principle. I'm less sure that the resulting pairwise protocol would always have desirable efficiency and robustness characteristics ... eg., wrt efficiency, compare true broadcast with a pairwise emulation; or wrt robustness, the pairwise protocol might depend on an auxiliary coordinator which might be a single point of failure. From a non-technical point of view, I'm worried that legal considerations might get in the way of reducing multi-party to pairwise interactions: in a jurisdiction which recognizes multi-party agreements as legal primitives it could be the case that a transformation to pairwise agreements changes the legal landscape in unexpected or unfortunate ways. Maybe there just aren't any jurisdictions which recognize primitive multi-party agreements, in which case this is a non-problem. But nobody has managed to provide me with convincing evidence that this is so (yet), and I can't think of any a priori reason why it should be. And even if the legalities are pairwise-friendly, it might still be the case that informal business agreements are often multi-party, with an only hazy connection with background legally binding pairwise agreements. Those informal agreements might Just Work sufficiently well in enough cases that any disconnect with the legal background aren't troublesome. On the contrary, aligning a practical multi-party agreement with a strictly legally binding set of pairwise agreements might be too time-consuming and expensive to be worthwhile. The problem here is that automating this stuff would be dependent on that alignment being done upfront, and that might be a significant obstacle. I for one don't know how to begin to answer these last questions ... this looks more like territory for lawyers and sociologists than for protocol designers. Cheers, Miles
Received on Sunday, 22 December 2002 07:06:01 UTC