[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Another suggestion for the BNF

• Subject: Re: Another suggestion for the BNF
• From: Nick Spadaccini <nick@xxxxxxxxxxxxx>
• Date: Tue, 2 Jul 2002 09:08:09 +0100 (BST)

On Mon, 1 Jul 2002, Herbert J. Bernstein wrote:

>   As a computer scientist, you must be aware that _no_ BNF can fully
> and accurate represent a context-sensitive language, such as CIF,
> and that users of any BNF for CIF (as opposed to a context-sensitive
> set of productions for CIF as we have disccuessed in the past) need
> very strong warnings to consult text notes and lexer suggestions
> before attempting to understand the BNF or to create a parser.

I think you are preaching to the preacher, Herb. The members of this list
will recall my repeated attempts to educate them that a BNF cannot fully
express the grammar of STAR or CIF and that in building a parser there
were many aspects that would need careful consideration.

>   Also, as far as I am aware, there is no BNF that has formally been
> adopted by COMCIFS are fully and accurately representing CIF.

I am trying to recall where I suggested that the BNF on my website was in
any way, shape or form the formally adopted definition of CIF. I apologize
to the list community if I have given the impression mine was more than
*a* BNF for CIF - albeit under constant review.

>   I have written Brian and Syd on this matter, and hope you, as
> a responsible computer scientist, will join me in urging caution.

Yes, anybody here who thinks they can they can take the BNF productions
and automatically crank out a parser that deals with the semantics is
dreaming. But I don't think anybody here does think you can do that with
the CIF BNF productions - or am I dreaming?

> Copy of message to Brian:
> 
> Dear Brian,
> 
>   Are there _any_ BNF's now posted as supposedly representing an agreed
> definition of CIF?  If so, this is very dangerous, and damaging to the
> future of CIF.  It appears that some people (e.g. Hester) are under the
> impression that some of Nick's sketchy and incomplete BNF's are formal,

Sorry to waste your time James. Clearly the productions I have written
seemingly bear little resemblance to the CIF grammar.

> complete agreed definitions of CIF.  That will result in the creation
> of bizarre dialects of CIF (it has already happened at least once).

Could those authors of bizarre CIF dialects please indicate to me what
aspect of my BNF has obfuscated their understanding. I will be happy to
assist them in their revisions.

>   Please, if the IUCr site has any BNF posted, please take it down, or at
> least post a warning that "no formal BNF for CIF has yet been agreed by
> COMCIFS, and those who create parsers from any BNF should be aware, that
> since CIF is a context sensitive language, _no_ BNF can possibly be a
> complete speficifcation of CIF, and all text notes and lexer suggestions
> should be read very carefully in creating a parser from any proposed BNF
> of CIF"

The IUCr could do this, but anybody with the slightest clue would already
be well aware of it. The IUCr would do a better service by not being so
patronizing to its reader base.


yours irresponsibly

Nick

--------------------------------
Dr N. Spadaccini                           Deputy Head of Department

Department of Computer Science &            voice: +(61 8) 9380 3452
Software Engineering                          fax: +(61 8) 9380 1089
University of Western Australia            email: nick@cs.uwa.edu.au            
35 Stirling Highway                      w3: www.cs.uwa.edu.au/~nick
CRAWLEY, Perth,  WA  6009             
AUSTRALIA                               CRICOS Provider Code: 00126G

Reply to: [list | sender only]

• Prev by Date: Re: Another suggestion for the BNF
• Next by Date: CIF parser / dialects (was Re: Another suggestion for the BNF)
• Prev by thread: Re: Another suggestion for the BNF
• Next by thread: Modification to BNF needed for <line_of_text>?
• Index(es):
  • Date
  • Thread