Re: Proposed Up/Down protocol description

[Geoff Huston <gih@apnic.net>]

At 12:35 PM 24/05/2007, Robert Loomans wrote:

> > It's possible to restrict the minimum and maximum length of otherwise
> > freeform strings with XML Schema Datatypes.  It is also relatively
> > trivial.  It is harder to agree on a sensible maximum length, since
> > there are so many reasonable choices that are driven primarily by
> > preference.
>
> Some of the strings have natural lengths:
>
> - type, status, code, etc are enumerated
> - g(SKI) and g(AKI) are an exact length
>
> (BTW, do we *mean* g(SKI), ie, g(SHA1(public key)), or do we really just
> want to put in the hex of SKI, ie, SHA1(public key), instead? The only
> interesting thing about g(SKI) was that it's filesystem/URL safe.)


SKI = sha1(public key) in my lexicon

so g(ski) is g(sha1(public key))





> - suggested_sia_head is a filesystem path so it should be some sort of
> lowest common denominator of the PATH_MAX of the platforms that will be
> running this (typically 1K)
> - cert serial length(decimal(2^(8*20))) (RFC 3280 says 20 octets)
> - msg_ref is length(decimal(2^64)) (or maybe length(decimal(2^128)))



> The remaining fields, sender, recipient, class name and the
> resource_set_*, we could take one of two approaches:
>
> - try to estimate what would be "sensible" maximums based on the
> conceivable maximum number of entities involved. eg, sender, recipient,
> and classname are 255 chars, resource_set_* are 4k
>
> - make the maximums stupidly large. eg, sender, recipient, and classname
> can be 1K (or 4K, or....), resource_set_* are 8k (or 16k, or....)


see my prev posting on this for an initial suggestion


g