[cvsnt] Re: Keyword 'properties' breaks rcs keyword ordering

[Tony Hoyle]

Bertho Stultiens wrote:
> Well, I do not know why this is done, but it is not a smart move IMO. It is
> so simple to adhere to the rcsfile(5) format, so that other applications can
> read the content without being broken.

The wrappers have been around for 2 years at least and were written to 
allow applications to interoperate.  I said back then and have always 
said that applications should not be making assumptions about the way 
cvsnt stores its data.

Backward compatibility disappeared sometime before binary deltas were 
implemented and hasn't been there since.

> Why should you want to require the cvsnt RCS wrappers? I can see no other

They've been required in most instances for a very long time.  Their 
purpose is to allow applications such as ViewCVS to work correctly 
(although ViewCVS doesn't use the wrappers any more.. it calls 'cvs 
rcsfile' directly these days).

> If you want to improve on the fileformat, then, I guess, you should redesign
> all ambiguities in the definitions of rcsfile(5) and move to a completely
> _new_ format. That would be understandable for outside developers (and
> easier to work with).

That's been happening for a long time.   It started as RCS, became 
'almost' RCS and is heading for something completely different in the 
future.

Tony