Re: ICC operating system identifiers

[Graeme Gill <graeme2 argyllcms com>]

Richard Hughes wrote:
> If you think it may be useful, then I guess "*nix" makes most sense as
> Graeme has been using this for ages. If anything, Graeme's email would
> convey more weight than mine ever would.

The issues are the following:

1)  The platform signature is not of great importance, and it's difficult
    to know what the original intention of it was. ICC profiles are
    typically not platform specific. It's lack of importance means
    that it might be difficult to excite anybodies interest in
    this issue. The only reason I added '*nix' to my code was an
    interest in not being misleading. My other alternatives were
    to use one of the other two Unix based platform signatures (SGI or
    Sun), or to set it to Unknown (0).

2)  Most of the other ICC signatures are in registries, so it's
    not that bigger deal to ask someone to add a signature to
    a registry - a single person will have the authority to
    just do it (hence 'argl' added as a CMM signature). The
    platform signatures only appear in the ICC spec. though,
    and they also seem to correlate with the founding members
    (Although I notice that Taligent has been dropped between V2 an V4).
    So to add a new signature means someone within the ICC has to
    propose a spec. change and it then has to be approved by the members
    at some point !

3)  I'm not sure how much consensus there is about a signature that
    would cover Linux. My thought process was that the existing
    signatures were both too vague and too specific. They
    are really operating system vendors, not specific operating system
    platforms, and so are fairly loose (there is a noticeable difference
    between Win3.1 and WinNT, between OS9 and OS X). They are also
    too specific, in that Sun and SGI were both Unix based systems,
    although SGI has switch to Windows, so what does it mean ?
    If the signatures were just a registry, there wouldn't be any
    problem being very specific (Linux, FreeBSD, OpenBSD, RedHatLinux, etc. etc.)
    So I went for something that would cover cover Linux, BSD etc. and would
    be enough to distinguish it from Apple and Microsoft.

As I suggested on the OpenICC list, if there is sufficient consensus
then I'm happy to mail a couple of people and request that a new platform
signature be put forward as a modification to the ICC spec., but on the
other hand I won't be surprised if action on it is a little slow :-)

cheers,
	Graeme Gill.