Re: java & endianness [Re: Binary tx format for an array?]

On Mon, 2006-07-10 at 19:54 +0200, Marc Herbert wrote:
> My point is that it IS universally true in the Java universe AND in
> the networking universe. There simply cannot be two opposed networking
> universes... that would mean we'd have two Internets for instance?
>
> So I think the whole point of Java is that everything is big-endian, so
> you do not need to know about endianness anymore. IMHO this is a
> success.

Just because Java is big-endian and this particular protocol is big-
endian does not mean that the universe is big-endian. There are
protocols which either always use little-endian, or allow the endian-
ness of parameters to be negotiated for performance. One example is the
TDS protocol, so you may want to look at the open source jTDS JDBC
driver and see how they handle it.

Your point is however well taken that there are more big-endian
protocols out there than little-endian protocols, and Java works easily
with those.

> Interesting. Looks like you need this only when dealing with
> badly-behaved applications that do not use the network order, right?
> Should be very seldom, don't you think?

It's not badly-behaving applications, it's protocols whose designers
chose to use little-endian formatting, usually for performance reasons
on common x86 hardware.

-- Mark