[Sis-csi] noting unusual CCSDS security critique

Scott Burleigh Scott.Burleigh at jpl.nasa.gov
Thu Oct 6 01:24:35 EDT 2005


Lloyd Wood wrote:
> On Wed, 5 Oct 2005, Scott Burleigh wrote:
> 
> 
>>Since code footprint size and increased performance are not aspects
>>of protocol design,
> 
> 
> certainly not for protocol designers who can't code (and are
> generally off in IETF workgroups designing the unimplementable.)

We seem to be getting off the point a bit here, but aren't you being a little harsh on IETF?  I think in the vast majority of cases the people in those work groups are capable programmers designing protocols and protocol features that are perfectly implementable.  Not necessarily useful in every case, but sometimes it's hard to know that until you build the thing and try it out.

> It's one thing to look at a protocol design specification and say
> 'this will likely be difficult/complex to implement' or 'this will
> perform poorly thanks to these design choices'.

Sure, anybody can do that, though sometimes people get this wrong too.

> It's rather harder to look at an implementation and determine that 'the size/performance of
> this implementation results directly from and depends solely on the
> underlying protocol design', however much you may want to.

But who would ever want to say anything so foolish?

My point was much more straightforward.  Protocol designs are typically expressed abstractly rather than in code written in any particular programming language.  Since they are not expressed in code, they don't compile; therefore they have no code footprint size and cannot execute, and since they cannot execute they have no performance profile.

Now, certainly some aspects of a protocol design (functional scope, variability, etc.) may tend to increase or decrease the chance that an implementation of the protocol in a given language will have a compiled code footprint that doesn't exceed some limit -- but of course the choice of language does this too, as do the skill of the programmer, the amount of time allocated to accomplishing the implementation, the choice of compiler, and any number of other factors.  The point is that it's the implementation that has a code footprint and a performance profile, not the protocol design.

We can talk about this some more if you want, Lloyd, but we should probably take it off this mailing list.  I suspect we've already exhausted most list members' patience for this topic.

Scott



More information about the Sis-CSI mailing list