Re: draft proposal for catalog resolution

In message <9703311912.AA08755@sqrex.sq.com> lee@sq.com writes:
> > > Record ends are just part of whitespace in the catalog.
> > 
> > Is there any good reason that we need that flexbility? Perl programming
> > really is a lot easier when things are one per line.

I can't speak for the other developers and implementors, but parsing CATALOGs
is not a major problem if the spec is clear.  (I've hacked a CATALOG parser
which managed nested catalogs and it seemed to work).  At least there aren't
any p*r*m*t*r *nt*t**s in them.  Record ends are just counted as whitespace.

> Agreed.
> Also, a common error seems to be omitting the double quotes round
> an argument to a CATALOG entry; if entries had to be separated by
> at least one of \n \r or \f, this requirement could be dropped, no?
> Otherwise given
>     STYLESHEET "Pretty little borders and teapots" TEAPOT

So long as the quotes are balanced (which will have to be the case for XML
attribute values anyway) there is no problem parsing this, even if it's
broken over several lines (so long as whitespace is folded into single spaces
which we seem to be converging on.)

> the TEAPOT is taken as starting a new entry, which is very confusing.

One problem with using record ends is that you soon start having to think
about continuation characters, etc.  Of course you can always put a character
in column 6 :-)


Peter Murray-Rust, domestic net connection
Virtual School of Molecular Sciences