🚀 go-pugleaf

Mike Girou (girou@parashift.com) wrote:
>Once [Microsoft] publish the internal [Winword file format]
>standards, then they are pretty well stuck with that standard for
>life.  As long as they, and only they, build code around the internal
>representations, they are free to change those representations as the
>need arises (hopefully providing conversion tools).  

Just like OOP!


gt2497b@acmez.gatech.edu (Joe Novosel) wrote

>Their own programs are not even compatable
>with themselves.  I use Word v6 and v2.  I have to use v2 on my notebook
>because 6.0 is so big.  I have to be very careful file formats because 2.0
>won't read 6.0's files.  

There's a Microsoft filter for Word 2 that reads Word 6 files; 
can't remember the name but I think you can ftp it from microsoft.com. 
Still doesn't support the nice new formatting features in WW6 of
course ... 


WinWord 6 file format:

WinWord 6+ and some or all of the rest of MS Office use OLE Structured 
Storage for their files. This means each file has an internal tree 
structure of _storages_ and _streams_, analogous to directories & files. 
It gives a standard interface for storage of OLE objects, and for other 
apps to extract information (e.g. summary info). It is also why WW6 files 
are so much bigger than earlier Word versions'.

There's a Delphi-centred intro to OLE SS by John Lam in PC Magazine
(US) 19 December 95, though it doesn't cover WW6 as such. Another
contributor to this thread mentioned discussion of WW6 files in PC
Mag; like him/her I recollect it but can't find it.

Word tradition from way back has been to store the text as straight
text with no formatting instructions or tags. All the formatting 
is in separate tables or lists at the end. This made it pretty easy
to extract plain text from a not-too-complicated Word doc: the text
began immediately after the file header, and the header contained a
pointer to the first byte of the formatting tables.

The same basic approach should work with WW6, although you'd have
to delve into OLE SS to get the offsets of the start and end of text.

With complicated texts (footnotes, annotations, symbols, tables, 
fields), however, things rapidly get a lot nastier. 

Footnotes have always been indicated by a single non-printing char 
in the text (same char for every footnote); the program had 
to keep track of these and maintain pointers to the text of each 
note. From WW6, the same technique is used for symbols (e.g. a single
char from WingDings or Symbol font) (which is why searching for a
symbol in WW6 doesn't work properly!).

IMHO the least worst approach to the problem this thread started 
with would be to use WordBasic to export the relevant files as text 
or .RTF and distribute them in that form. No big challenge to write 
a WordBasic program to look out for changes (SaveDate) in the docs 
and update the text or RTF versions (or you can do it with Delphi 
and the WordBasic API).


-- 
Best wishes

John Nurick

e-mail: j.nurick@dial.pipex.com 
v-mail: <+44|0> 191 281 1306