Re: Free UNIX for non-commerical use.
From: Dr. David Kirkby (drkirkby_at_ntlworld.com)
Date: Fri, 08 Aug 2003 10:37:09 +0100
Tony 'Nicoya' Mantler wrote:
> In article <3F330255.B90F491B@ntlworld.com>,
> "Dr. David Kirkby" <email@example.com> wrote:
> : The problem is that a high
> : percentage of developers develop code on a machine which is
> : a) 32-bit
> : b) big-endian
> x86 is little-endian
Sorry it was late and I did not even think about what way around it
was. Yes, I accept that the bog-standard PC, which is what I was
meaning, is little-endian.
In my life I've even come across a machine which I guess you would
call mid-endian or something like that. It was a 24-bit machine, with
either most (or might have been least) significant byte in the
middle!! That was on a Bruker NMR (or MRI as it's now called) clinical
> Thanks primarily to Debian, endian errors are getting very rare, and
> pointer-size errors are slowly on their way out too.
I don't know what Debian's contribution is, but it can't get around
the fact that if people write binary data, it will be unreadable on
another system, unless they make provision by having a switch to read
data in the other format, or they did what I once did, which was to
write a program such that the binary data was written in one format,
irrespective of the platform.
> : I'd like to read the temperature of my SCSI disks. There are several
> : Linux utilities to do this, but not one will work on SPARC. Yet the
> : author of 'cdrecord' has clearly demonstrated it's possible to write
> : low-level SCSI code that is portable.
> Possible but difficult. Every single platform uses a different API to send
> commands to and from drives. cdrecord gets around this by (no surprise) writing
> different drive interface code for every single platform.
In which case, if you develop a program that needs low-level SCSI, why
not start from Jörg Schilling's code, where the API is I assume
consistent for cd-record, and the interface code is platform specific?
Jörg Schilling might like to comment on that.
I accept low-level code is more difficult, but if people had access to
different systems cheaply, they might think about these issues more.
But there are many bits of code around where the too much is assumed
about the platform, that does not need to be. It might assume 'memset'
exists, which is not always the case. It might assume 'long long'
exists, which is not always the case. I can't help feeling that making
things like the OS and development tools readily available for
non-commercial use, will encourage more people to look to use other
hardware, and perhaps realise gcc is not the only compiler in the
world, and some of gcc's options are not accepted by someone else's
> One of these days I might port smartmontools or some other package to IRIX, but
> I don't have a Sun so you're out of luck there.
Well I have an SGI too, but are less concerned about that, as it only
runs a few hours/week and don't bother me too much. Yesterday and
Wednesday I had a lot of temperature problems on my older SPARC 20's
which have fast disks and fast CPUs in them. Hence my 'brute force'
approach to over-temperature protection
came into effect.
-- Dr. David Kirkby, Senior Research Fellow, Department of Medical Physics, University College London, 11-20 Capper St, London, WC1E 6JA. Tel: 020 7679 6408 Fax: 020 7679 6269 Internal telephone: ext 46408 e-mail firstname.lastname@example.org