Re: Refactor of ixml grammar

My only argument is that as a general design principle, it is better
to be able to specify a given behavior or option by writing something
than to be able to specify it only by writing nothing.  I think that that 
principle is a good one even in cases where in practice very few people
need or want to make use of that ability, or none do.  So I can agree
wholeheartedly that I cannot now imagine that anyone will actually feel 
a need to write ^ in front a terminal, and nevertheless say that if 
someone does feel that need, I would like them to be able to write
^ in front of a terminal.

I recognize that few people will find this line of thinking persuasive
unless they are already convinced of the conclusion.  If I could think
of an argument more likely to persuade you, I would bring it forward.

Michael


> On 6,Nov2021, at 8:45 AM, Steven Pemberton <steven.pemberton@cwi.nl> wrote:
> 
> Well the default for a terminal is to include it in the output. So
> 
> 	date: day, " ", month, " ", year.
> and
> 	date: day, ^" ", month, ^" ", year.
> 
> are equivalent. You need ^ for nonterminals because their default can be hidden or attribute, and so you need to be able to override it. But there is nothing to override for terminals, and I can't imagine anyone ever feeling the need to use the ^ form for terminals.
> 
> Steven
> 
> On Thu, 04 Nov 2021 15:15:20 +0100, C. M. Sperberg-McQueen <cmsmcq@blackmesatech.com> wrote:
> 
> > Oh, er, ah. I am always uncomfortable when there is no explicit
> > specification whose meaning is the default behavior.
> >
> > (An early influence on me, I guess, was an article by Harlan Mills of
> > IBM on how to use System 360 JCL with fewer errors; the basic idea
> > was that many errors were caused by the complicated system of
> > defaults, which changed depending on various partial specifications.
> > His remedy: write JCL as if it had no default values. That way, you
> > never needed to learn the rules for defaults. I do still do that when
> > learning new languages; it helps. I also decline to learn more than
> > one or two levels of operator priorities.)
> >
> > Michael
> >
> >
> >> On 4,Nov2021, at 4:34 AM, Steven Pemberton <steven.pemberton@cwi.nl>
> >> wrote:
> >>
> >> Another change I think worth making:
> >> A tmark is either ^ and -, since @ is completely meaningless for
> >> terminals.
> >>
> >> However, ^ also has no use, since that is the default anyway, and there
> >> is no usecase where you need it to override anything.
> >> So I propose making:
> >>
> >> @tmark: "-".
> >>
> >> Steven
> >>
> >> On Thursday 04 November 2021 00:02:54 (+01:00), Steven Pemberton wrote:
> >>
> >> > In an idle moment, I refactored the grammar. Comments gladly received.
> >> > Changes: * I hid all nonessential terminals. I know above all Tom was
> >> asking for this.
> >> > * I moved the spaces from the rule for ixml into the rule for rule.
> >> Tidier and more consistent.
> >> > * I renamed S to s.
> >> > * I simplified 'namestart', since I realised class L covered all the
> >> cases.
> >> >
> >> > I think that's all.
> >> >
> >> > See attachment.
> >> >
> >> > Steven
> >>
> >
> 
> On Thursday 04 November 2021 15:15:20 (+01:00), C. M. Sperberg-McQueen wrote:
> 
> > Oh, er, ah. I am always uncomfortable when there is no explicit
> > specification whose meaning is the default behavior.
> >
> > (An early influence on me, I guess, was an article by Harlan Mills of
> > IBM on how to use System 360 JCL with fewer errors; the basic idea
> > was that many errors were caused by the complicated system of
> > defaults, which changed depending on various partial specifications.
> > His remedy: write JCL as if it had no default values. That way, you
> > never needed to learn the rules for defaults. I do still do that when
> > learning new languages; it helps. I also decline to learn more than
> > one or two levels of operator priorities.)
> >
> > Michael
> >
> >
> > > On 4,Nov2021, at 4:34 AM, Steven Pemberton <steven.pemberton@cwi.nl> wrote:
> > >
> > > Another change I think worth making:
> > > A tmark is either ^ and -, since @ is completely meaningless for terminals.
> > >
> > > However, ^ also has no use, since that is the default anyway, and there is no usecase where you need it to override anything.
> > > So I propose making:
> > >
> > > @tmark: "-".
> > >
> > > Steven
> > >
> > > On Thursday 04 November 2021 00:02:54 (+01:00), Steven Pemberton wrote:
> > >
> > > > In an idle moment, I refactored the grammar. Comments gladly received.
> > > > Changes: * I hid all nonessential terminals. I know above all Tom was asking for this.
> > > > * I moved the spaces from the rule for ixml into the rule for rule. Tidier and more consistent.
> > > > * I renamed S to s.
> > > > * I simplified 'namestart', since I realised class L covered all the cases.
> > > >
> > > > I think that's all.
> > > >
> > > > See attachment.
> > > >
> > > > Steven
> > >
> >
> >
> >
>

Received on Saturday, 6 November 2021 17:51:42 UTC