Hi Michael,

Michael Richardson wrote:

>I hope that the compression of DNs will be distinguished such that there is
>only one way to do it, and one can compare the compressed forms rather than
>having to expand it to be sure.
>I don't feel confident that this will be the case because I think that there
>will always be an "out" where raw DER bits can be included for things that
>the compressor doesn't know how to deal with.  So a stupid compressor might
>produce a bigger result than a smarter compressor.

A long as DER is Distinguished (which it should be) and the CBOR encoding is 
one-to-one (which it should be) comparing the CBOR encoded DNs will not be a 
problem.

It is hard to comment unless you give more details on why you believe the CBOR 
encoding would not be one-to-one. 

As long as certificates really are DER, I think everything will work. If some 
certs are BER in disguise, we will get problem. 

>I also wonder about this work in the context of Packed CBOR.
>Should we restart our thinking assuming that we are really just building a
>static directory for Packed CBOR?  I suspect that this bespoke compression
>will do better for bytes-on-the wire, but what is the code size impact,
>assuming that Packed CBOR code was already available?

I do not see how Packet CBOR would replace anything in this context. What is 
mainly needed here is a DER -> CBOR  one-to-one function. CBOR packing is CBOR 
-> CBOR and it is not one-to-one (it is not a function). Packed CBOR seems to 
work best on very structured data. Packed CBOR with or without a static 
dictionary could be used to pack a c5chain of CBOR certificates. I guess you 
could also theoretically combine a different (not yet specified) DER -> CBOR 
one-to-one function with a (not yet specified) deterministic one-to-one CBOR 
packing, and a (not yet specified) dictionary, but I think the "compression" 
would be limited.  


_______________________________________________
COSE mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/cose

Reply via email to