[oc2e] Serialization?

Alyssa Milburn wrote:
> On Mon, 09 Jan 2006 14:50:27 +0100, Thomas J. Boschloo wrote:
>
>>And saving 16 bit words to disk in an intel environment will result in
>>byte swapping, which is exactly what you don't want here if you follow
>>RFC 1014!
>
>
> It won't result in byte swapping, since that's how it is inside the
> processor, and in memory, etc, it'll just result in a different endianness
> to the big-endian 'standard', which you might want to fix by byte swapping.
>
> The thing is, I'm not convinced of the point of RFC 1014, it could be
> summarized as saying "write the raw standard numbers, making sure that the
> hardware copes with byte endianness, and you swap the data to big-endian
> if necessary".

I would swap the octets at all times. It is just that on a G4 processor
the function you use does absolutely nothing. But on a Pentium II it
becomes active due to a compiler option.

At least that is how I learned it..
You want your saved norns to be exportable to all platforms.. Not just
Intel And you do need two different processors (in the endian way)
to test if your code works..

Thomas
--
Gothika: "How can you trust someone who thinks you are crazy"

On Mon, 09 Jan 2006 15:02:20 +0100, Thomas J. Boschloo wrote:
>> The thing is, I'm not convinced of the point of RFC 1014, it could be
>> summarized as saying "write the raw standard numbers, making sure that the
>> hardware copes with byte endianness, and you swap the data to big-endian
>> if necessary".
>
> I would swap the octets at all times. It is just that on a G4 processor
> the function you use does absolutely nothing. But on a Pentium II it
> becomes active due to a compiler option.

I suppose. But that's only useful works when your target endian format is
big-endian, which is getting a lot less popular, and so will only be
useful for our own unique data files.

Anyway, the serialization code we're using does all the endian stuff
automatically. The most efficient way to do it is probably to always use
native endianism, though, and note at the start of the file which
endianism it was saved using - because people are going to be
importing/exporting stuff on their own computers far more often than
they're going to be sharing them with people using other endianisms. The
X11 protocol does that, for example.

> You want your saved norns to be exportable to all platforms.. Not just
> Intel And you do need two different processors (in the endian way)
> to test if your code works..

I'm on powerpc, which is big-endian. Everyone else involved is on
little-endian, pretty much. All the Creatures data files are little-endian
ones because they coded it on x86 and gave no thought to portability. So I
know how annoying endian issues are, trust me. It shall be tested.