Re: Compromise on canonicalization

[Danny Ayers <danny.ayers@xxxxxxxxx>]

On Thu, 2 Sep 2004 12:39:10 -0700 (PDT), Dare Obasanjo <kpako@xxxxxxxxx> wrote:
> 
> --- Danny Ayers <danny.ayers@xxxxxxxxx> wrote:
> >
> > +1 to MUST use URIs as ids
> > +1 to consumers/republishers MUST maintain the ids
> > char-by-char
> > +1 to id generators (publishers) MAY canonicalize
> > URIs
> > +1 to id comparators (consumers/republishers) MAY
> > use char-by-char comparison
> 
> -10
> 
> Why the heck are you dragging this thread out? Paul
> gave a great description of the consensus on the list.
> You're above description is a giant leap back.

My estimate of opinion on each is:

URIs as id  - strongly pro
maintain ids - strongly pro
publishers canonicalize - locked
char-by-char at client - inconclusive

Hence my stepping back on the third. Based on the technical arguments
put forward the position I suggested on the last point seems
reasonable.

> "id comparators (consumers/republishers) MAY
>  use char-by-char comparison" ?
> 
> MAY? WTF? Have you misread everything that has been
> posted about why existing specs like RDF and XML
> namespaces use char by char comparison? 

What I've seen on that specifically is 1. that they use it, and 2.
that not everyone's convinced it's done well, quoting Roy Fielding:
[[
It is just poorly phrased in the RDF specification.  RDF can define
how RDF does comparisons of URI references in the processing of
assertions.  However, that definition does not change the fact that
two URI that are equivalent (by their very nature) may be falsely
determined to be different by the RDF algorithm -- that is the nature
of lossy optimizations.  This is not a problem when the URI references
are provided in a reasonably canonical form.
]]


Have you
> missed statements by different aggregator authors
> pointing out that the main issue is not the comparison
> algorithm?

If the main issue is not the comparison algorithm, then why are you
bothered about char-by-char comparison being a MAY?

> Canonicalization buys you jack. The compromise was to
> have some wording in the spec saying its a good idea
> so stonewallers like you who seem determined to impede
> progress of the working group can let us move on to
> other issues. Instead you have not only stuck to your
> obsession with requiring canonicalization despite a
> number of examples showing its ineffectiveness in the
> real world as opposed to your contrived examples[0]
> but now are trying to open discussion on another
> consensus issue.

1. You will note that I changed to MAY on canonicalization above.
Hardly stonewalling.
2. Your examples [0] deal with *different* URIs

> Hell, no.

If you have convincing arguments why comparison at client ought to be
a SHOULD, let's hear them.

Cheers,
Danny.

> [0] http://www.imc.org/atom-syntax/mail-archive/msg09211.html
> 
> 
> 
> =====
> THINGS TO DO IF I BECOME AN EVIL OVERLORD #23
> I will keep a special cache of low-tech weapons and train my troops in their use. That way -- even if the heroes manage to neutralize my power generator and/or render the standard-issue energy weapons useless -- my troops will not be overrun by a handful of savages armed with spears and rocks.
> 
> _______________________________
> Do you Yahoo!?
> Win 1 of 4,000 free domain names from Yahoo! Enter now.
> http://promotions.yahoo.com/goldrush
> 


-- 

http://dannyayers.com