Re: iSCSI-related conclusions from Orlando interim Meeting

[Glen Turner <glen.turner@aarnet.edu.au>]

julian_satran@il.ibm.com wrote:
> 
> I think that David (Robinson) has a good point. I think that NDT would want
> to say text encoding in one of the C-language forms (0x'...' etc.) and those
> include binaries etc.

Hi Julo,

Perhaps not the C method, as this has caused problems in the past
(most notoriously with scanning IP addresses).  The difference between
"10" and "010" isn't apparent to the typical author of a configuraton
file (one number is in decimal, the other number is in octal).
People that align strings by zero-filling them are burnt.

If using UTF-8 you may also want to constrain valid strings to the
shortest UTF-8 representation of the string[1].  This makes checking
for values much simpler and allows the writing of reliable regular
expressions:

  eg:  allow volumes in "*.example.au"

> The alternative is to allow only hex (not always nice nor readable).

Or decimal.  Probably better if values will propogate up into
configuration files.  One of the advantages of a textual representation
of numbers is that you can simply pass down the user's parameters
unaltered.  This allows additional parameters to be passed to the
server without the client needing to be updated.

Regards,
	Glen

[1]  I can't find the reference for the right word for
     "shortest", it might be a word like "canonical"
     instead?