[vorbis-dev] Xiph Magic

[J C Fitzgerald]

On Fri, Jul 18, 2003 at 02:08:56PM +0300, Beni Cherniavsky wrote:
>
> libvorbis has no ambiguities [...] yet it's not good as a spec
> because it's hard to know that what it specifies is what Monty
> really had in mind (although it must be very close by now, all
> big bugs have been fixed already) and more importantly, it is
> not easy to read and to create new implementations of it.
>
If you mean the code of libvorbis then it may also exhibit different
behaviours on different platforms if it uses any constructs the language
standards committee deemed 'unspecified', 'implementation defined',
'locale specific' or 'undefined'.  By it's nature, it no doubt also
contains overspecifications.

And how can you tell objectively if all big bugs are out unless you have
a specification to test against?  I've had software in the past which
has undergone similar bug-squishing refinements until it was deemed to
be major-bug free only to have a real big bug crop up quite unexpectedly
some years later.

>
> That's why formal specification languages
> are not very popular, as far as I see - they are not convenient for
> importing into a human's mind.
>
But code isn't a formal specification; it isn't precise enough (strange
as that may sound).  Formal specification languages are actually easier
than informal ones to assimilate because they have a powerful notation.
Studies suggest that even users prefer their manuals to include formally
specified behaviours in them because they can unambiguosly determine how
any given operation should behave.  (And learning to read the language
isn't particularly difficult.)

> 
> Yet some specifications, most notably RFCs, while written in informal
> natural language, are very reable and leave remarkably little
> ambiguity when read by a human being.
>
When reading them with some formal methods experience, I certainly
notice ambiguities.

>
> So I don't understand what's your precise point here.
> 
I was simply expressing the POV that we shouldn't be surprised if
people have different interpretations of an informal specification.
I imagine that, like me, most people have only read the bits that
answered specific queries they had and not the whole thing.  So the
different interpretations may arise for a number of reasons.

I wasn't suggesting that specification be rewritten formally, just
that we accept the limitations of an informal specification.  And
better forums exist for litigating the desirability of formality
(but we can discuss it privately if you wish).

John
--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.