Subject: Re: style => name?
From: John L. Clark (jlc6@po.cwru.edu)
Date: Sat Sep 08 2001 - 15:10:55 CDT
On Sat, Sep 08, 2001 at 12:32:14PM -0700, Paul Rohr wrote:
> At 10:34 AM 9/7/01 +1000, Martin Sevior wrote:
> >OK, I cleaned this up. For some reason that I don't understand, sometimes
> >abi was using the attribute "name" to repesent a style and sometimes it
> >was using "style" to repesent a style.
>
> Sorry. I thought the original implementation was pretty easy to understand.
> There are two ways styles are used in a document:
>
> 1. All style *references* take the form of a STYLE= attribute on a P or C
> tag. The semantics are that you're applying a paragraph or character-level
> style with that name (as appropriate) to the current scope.
>
> 2. All style *definitions* are S tags inside the STYLES block at the head
> of the document, where the two immediately relevant attributes tell you what
> the NAME= of that style is and what TYPE= of style it is (P for paragraph or
> C for character).
This is exactly as I understood it, and it made (and still makes) sense
to me. I spoke with Martin in irc a couple days back, however, and he
said that he found "name=..." properties in paragraphs in some of his
old documents, although I couldn't confirm that in any of the documents
I have.
He claimed that it was a bug, and that it was simply harder to convert
"everything" to 'style' properties than it was to go to 'name'
properties.
> Now that I've explained, would you mind reverting to the original file
> format? Using a NAME= attribute on paragraphs to indicate which style is
> being used is non-intuitive, and having all that compatibility code to
> maintain sounds bug-prone.
I strongly second this (+1) - I think it was just fine the way it was,
and if there was any spurious 'name=...' attributes anywhere in blocks,
then those should simply be fixed to be 'style=...'.
> Unfortunately, that DTD was out of date soon after Sam wrote it
> because none of the active developers knows enough about DTDs to keep it in
> sync.
I may have to take a look at that - I've been doing a lot of DTD
research lately, and I believe I could keep it up to date. That might
be helpful.
> As explained above, I guess I don't understand why we'd need to choose one
> or the other, since reverting to the original behavior should work just
> fine.
>
> Or am I missing something here?
Even after discussing this with Martin directly, I still have to agree
with Paul - I don't understand the original problem which prompted the
change.
And later, he wrote:
> Ick. It just occured to me that this file format change for style
> references (from STYLE to NAME) sounds like a showstopper for 0.9.3.
Indeed.
> In the future, we need to be a *lot* more explicit about any file format
> changes which might break compatibility.
Any file format changes /at all/, I would think. The dev list should be
informed of the status of the file format.
> motto -- make haste slowly
I couldn't agree more.
Take care,
John
This archive was generated by hypermail 2b25 : Sat Sep 08 2001 - 15:12:26 CDT