Re: [Cfrg] A way out of the little/big endian dispute?

[Phillip Hallam-Baker <[e-mail]>]

On Mon, Mar 30, 2015 at 9:59 AM, Watson Ladd <[e-mail]> wrote:

> Perhaps its
> because you've never been able to point to code

Watson, I am almost 50. I have been earning a full time wage as a
professional programmer since I was 14 and it is now several decades
since I handed in my doctoral thesis.

In what universe do you think you can win that game?

This particular topic was being discussed in IETF before you were born:

https://www.ietf.org/rfc/ien/ien137.txt


> in SSH that is more
> complicated because of the use of little-endian in Curve25519, or
> explain why a convention for the representation of short lengths in
> network protocols has to determine the representation of opaque values
> (which, as MD5 shows, it doesn't).

My concern isn't the complexity of code I am not going to have to
write, it is the fact that this proposal has to pass multiple levels
of IETF review and there can be no confidence in the outcome until the
final IESG sign off.

I know that if the TLS WG chooses big-endian there is absolutely no
possibility of the decision being overriden at a higher level and so I
can start telling folk to implement stuff the minute that there is a
spec.

Given that there is already TLS code and a spec, the scope for an
overturn there is quite small. But that argument does not carry over
to PKIX because there isn't PKIX code for Curve25519.

A proposal that attempts to re-open the network byte order decision
taken long before the IETF existed is almost certain to get a DISCUSS
vote and there is a very high probability of an appeal.


> The whole issue is writing the function that computes the scalarmult.
> You've never explained why this is made more difficult by
> little-endian, while it's clear that parsing half of all fields in a
> packet as little-enedian and half as big-endian is a pain. But it's
> very clear that defining two functions with the same name is a bad
> idea.

Umm, my code that does bit packing has no connection whatsoever to my
scalarmult code. I cannot conceive of a situation in which I would
want the two to have any relationship whatsoever. It is called
structured programming.