I would think the major reason Microsoft insists NT run only on one endianness
(which happens to be little) is transparent data-sharing: imagine a system
such as a VMScluster (soon to be ported to NT) in which every system can read
all the disks. How do you know that the data your application is reading is
little- or big-endian? You run into a lot of conversion issues, which (at the
very least) will eat into I/O performance. Naw, it's not worth the hassle -
every decent processor can be switched at boot time anyways.
Jan
The hassles have been overcome in the past. Sun's xdr library has been around
since 1986 or before. I recall having Motorola, SPARC and Intel based systems
all talking together via NFS in a seamless fashion. The xdr library took care
of the external data representations.
Regards,
Lew Wolfgang
>In article <JAN.95De...@cora.neuroinformatik.ruhr-uni-bochum.de>,
>Jan Vorbrueggen <j...@neuroinformatik.ruhr-uni-bochum.de> wrote:
>>
>>I would think the major reason Microsoft insists NT run only on one endianness
>>(which happens to be little) is transparent data-sharing: imagine a system
>>such as a VMScluster (soon to be ported to NT) in which every system can read
>>all the disks. How do you know that the data your application is reading is
>>little- or big-endian?
>The hassles have been overcome in the past. Sun's xdr library has been around
>since 1986 or before. I recall having Motorola, SPARC and Intel based systems
>all talking together via NFS in a seamless fashion.
Yep, I'm running NetBSD on Amigas and peecees and they can all talk to each
other and to Sparcs running SunOS. No problems at all.
-jm
--
I'm not lazy - I'm only selectively motivated
---> http://www.jmp.fi/~jmarin/ <---