Document: draft-ietf-idr-bgp-ext-communities-08.tx Review: Mark Allman Date: 17 februari 2005 This document is not ready, IMO. Comments: * I hate documents like this. The "what" is pretty well specified, but the "why" is basically untouched. This is a 14 page document and all that can be mustered is a one sentence abstract? And, the one-sentence really says nothing. The intro isn't much better. I would like to see a quick sketch of the issue that is being addressed would be a really nice addition. I.e., why aren't communities enough? I am not a fan of problem statements in general and I am not suggesting a pages long treatment, but a paragraph or two here there says why this document exists would be nice. * In section 2 the document says that types can be one or two octets, but makes no mention of how to determine if the type is one or two octets. I got it by the end of the document, but it seems like this should be explained more up-front. * Again, the "why" category of things, it would be nice to provide a quick example of the utility of passing the given information for each type of extended community defined. * It's not clear to me why there are two choices for values in the high-order type field. I am sure there is a good reason. Please spell out that reason. * I didn't quite get the distinction between the communities defined in section 3 (multiple definitions) with the definitions in 4--6 (one definition each). What is the distinction being drawn here? Also, you give diagrams of the format in section 3, but not in 4--6. Why? I think it should be consistent. * In section 6 the document suddenly starts to use standards language (MUST, MAY, etc.). It seems like this bit of rigor is missing from the previous sections. * [kompella-zinin] is referenced, but is not in the reference list. * Security considerations: Isn't the information conveyed potentially sensative (especially non-transitive information)? I think it would be good to note this. * It would be nice to have some contact information for the first author. An email address?