[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Last Call issues on USEPRO
On 6 Nov 2008, at 10:11, Russ Allbery wrote:
**#1583** USEPRO LC 5.2.3: Checkgroups control messages
There are several individual questions here.
One is handling of subhierarchies in combination with groups with
name as the hierarchy. I generally agree with Charles here that there
isn't a better way of handling the edge case than what we currently
and the group at the top of the sub-hierarchy will have to be listed
the parent checkgroups.
Sounds workable to me. Not the nicest solution, but I can't think of
any more correct way to do it.
Another is the serial number. I think there are two viable options
unlimited-length serial numbers (the current text) or sequence-space
arithmetic. I lean mildly towards the current text just because I
want to have to write the text for sequence-space arithmetic, and I
persuasive the argument that unlimited-length numbers can be checked
ASCII comparison and don't need to be treated as numbers.
That was my argument. :-) ( strlen(a, b) || strcmp(a, b) ) is
perfectly sensible as a serial number comparator, assuming trimming of
any leading zeroes and whitespace and validation that the rest of the
string is digits. Not sure if we have some kind of notes-to-the-
implementor, but that may be worth putting in as such a note.
Another is what should be done with a checkgroups that doesn't have a
serial number when checkgroups for that hierarchy previously did
serial number. I think this is a gap in the current specification.
Once you've accepted a serial number checkgroups, I think that you're
stuck with serial numbers, and non-serial checkgroups should be
ignored. Anything else seems to me to suffer the same issue as reset
messages below - there's no real way to definitively determine that
you're somehow supposed to stop caring about serial numbers.
I don't know if any extra wording is required to achieve that. If so,
it should probably be added.
The last is whether there's a means to reset the serial number.
Not really an issue if you go with either unlimited length or sequence
space arithmetic as the options for serial number. Neither of them
require a means to reset the serial number - and in fact in both cases
it's most likely counterproductive to have such a mechanism, since
there is no way to guarantee that any reset message is delivered.
It may be worth splitting this point into four issues; the first and
last may already be resolved, depending on whether anyone wants to
advocate doing something different about them than what we have now.
Well, my comments above. :-) I have nothing to say on the other issues.
~/ For those who've come across the seas, ~/ We've boundless plains to
~/ With courage let us all combine, ~/ To Advance Australia
-- The end of the second verse of the Australian National