This fundamentally breaks portability. If you are compliant with a
specification, then any feature you provide must conform to the
specification for that feature. This is essential for anyone attempting to
write a portable-application-using-extensions. Making the feature optional
doesn't change this. There are only two choices:
1) Remove the feature from the spec. This means that even
portable-application-using-extensions can't use it and remain conforming.
2) Declare as non-conforming any implementation that provides the feature
in a way that is not compliant with the spec. That means that GNU is
non-conforming.
/glen
-----Original Message-----
From: Andrew Josey [mailto:yyyyyy@xxxxxxxxxxxxxxxxx]
Sent: June 11, 1999 12:07 PM
To: yyyyyyyyyyyy@xxxxxxxxxxxxx
Subject: Re: Bug in XCU
Note: redirected to the austin-group reflector for general discussion.
A couple of notes that may be helpful when considering whether
to accept the suggested action, modify the suggested action or
reject.
(1) The "-z recsz" text has not been added to the POSIX semantics to
sort - it is a shaded extension over POSIX from the Single
UNIX Specification [marked as an XSI extension] (i.e. strictly
conforming POSIX systems need not support the feature). Hence no changes
noted in the CH. Its been present since the XPG3 definition of sort (1989)
(I do not have XPG2 on me to check further back).
(2) The text shading is marked as UN , which is a possibly unsupportable
XSI extension, which essentially means it optional.
(3) We have an action item from the previous meeting
to review all UN, PI , OF, OP coded text. Its possible based on the
meaning of UN that its presence may not be that useful to portable
applications.
It would be very useful if someone would take the action item in 3
and produce suggested aardvark for this draft, otherwise it will be on the
todo list for the evenings in Montreal:-)
best
regards
Andrew
On Jun 10, 10:55pm in "Bug in XCU", yyyyyy@xxxxxxxxxxx wrote:
> Bug report from : Paul Eggert , Twin Sun
>
> @ page 721 line 27297 section sort objection {2}
>
> Problem:
>
> The option `-z recsz' has been added to sort. This option should
> be removed, for several reasons.
>
> * `-z recsz' is incompatible with GNU sort, which
> uses -z to indicate that the input lines are terminated by NUL
> bytes instead of newlines. GNU sort's use of -z is more useful
> (and more used in practice) than `-z recsz'.
>
> * This option corresponds to a particular implementation of `sort';
> it is inappropriate for an implementation-independent spec.
>
> * The rationale (page 725 lines 27478-27481) indicates that
> `sort -z recsz' was omitted for good reason.
>
> * I see no justification for adding -z recsz. The CHANGE HISTORY
> (lines 27501-27504) does not mention it.
>
>
> Action:
>
> Remove the description of -z recsz
> (lines 27297-27302), and remove -z recsz from the synopsis
> (lines 27271-27273).
>
>-- End of excerpt from yyyyyy@xxxxxxxxxxx
-----
Andrew Josey The Open Group
Austin Group Chair Apex Plaza,Forbury Road,
Email: yyyyyyy@xxxxxxxxxxxxx Reading,Berks.RG1 1AX,England
Tel: +44 118 9508311 ext 2250 Fax: +44 118 9500110
|