On 3/23/19 4:41 PM, Michael Richardson wrote:
Thank you for this work.
It is very interesting description.
I would have liked to separate out the definition of the wire image a bit
From the value judgement that having a smaller wire-image is a good thing.
I read the document wondering if there was going to be a abstraction to
document what the wire-image of a protocol is. I saw no such thing, and
probably such an effort might result in attempts to boil an ocean.
I then wondered as I read the document if some kind of "WIRE IMAGE
CONSIDERATIONS" section might be suggested. I have put "PRIVACY
CONSIDERATIONS" headings in a few documents, and I know others have too.
I think that the wire-image description could fit into such a section.
Michael,
I think it would be useful for documents to specify which information
they expose; intentionally expose as well as what isn't intentional but
they need to expose to function. (For example, I believe IPsec ESP needs
to expose an SPI number to function.)
[And if it was me I'd put that information in the security
considerations section, but let's not discuss the set of sections in the
document and focus on the information in the document instead.]
But as the document mentions, the extent of the wire image is larger
than a single protocol (with the example being the DNS lookup(s) before
the main communication starts.) I wouldn't expect that system-level view
to be captured in each RFC.
Regards,
Erik
