At Mon, 7 Jun 2010 17:35:23 -0700, Terry Manderson wrote: > > Before I lend my support to either the simpler format or the current > compound TA format I'd like to see a pros/cons list comparing the two > approaches. > > While I find myself liking simple approaches but I also wonder if it > has any downfalls in use.
Terry, As far as I know the simple approach has no downside. We've been using something semantically identical to this in our testbed since last July, and the new syntax was a trivial drop-in replacement. The self-signed RPKI certificates change weekly as we get new data, validator keeps humming along. Other than simplicity for its own sake, the main advantage I see with the simple approach is that in some respects it is technically more robust, precisely because there are fewer moving parts, therefore fewer things to go wrong. This is not meant as a slam at the other mechanism, it's just that X.509 and CMS have a lot of options, each with multiple settings. The set of allowed options is much smaller than the set of possible options. Unchecked options can potentially trigger weird behavior in the underlying X.509 and CMS implementations, so all of the options must be checked. All this work makes no progress towards the real goal (verifying RPKI certificates), it's essentially busywork that one is stuck with doing because of the choice of trust anchor wrapper format. The simpler approach eschews most of this, and has very little to check beyond the checks that one has to do anyway for the self-signed RPKI certificate itself. So, to me, the simple approach adds exactly what we need beyond the base RPKI mechanism, and no more, while the other mechanism adds a lot of stuff that accomplishes the same purpose but takes more effort to get right. I freely grant that there are many other parts of this system which are also complicated and easy to get wrong, but that, in and of itself, seems insufficient reason to add another one. So, to me, at least in this case, simplicity is not just an esthetic virtue, it's a form of engineering robustness. YMMV. --Rob _______________________________________________ sidr mailing list [email protected] https://www.ietf.org/mailman/listinfo/sidr
