On 11/12/06, Gordon Sim <[EMAIL PROTECTED]> wrote:
>
>
>
>
>
>
> Steve Jones wrote:
>  >
>  >
>  > On 08/12/06, Gordon Sim <[EMAIL PROTECTED]
>  > <mailto:gordon.r.sim%40gmail.com>> wrote:
>  >  >
>  >  > For an opaque identifier, any
>  >  > 'meaning' which may be encoded in it (beyond its use as a unique
>  >  > identifier) is irrelevant (and in any case not visible) to the system or
>  >  > model to which it is opaque. It is not forbidden that some meaning may
>  >  > be discernible by some other system or model (e.g. the english 
> language).
>  >
>  > Not forbidden is a strange phrase. Naming conventions in programming
>  > languages almost always mandate that names _always_ be sensible and
>  > convey meaning.
>
>  REST is not a naming convention for a programming language though.

But the same principles applies as the names refer to resources etc.

>
>  > Not forbidding something means that it can't be
>  > relied upon to be true.
>
>  Why would I need to 'rely' on a URI having no meaning (to anyone, in any
>  language)?

Because if I see a URI of http://www.bar.com/foo/invoice then I'd
think that it referred to an invoice, unless I knew that URIs never
mean anything and therefore its just a random set of letters.

People make assumptions, names are a great way to get people to make
assumptions.

>
>  >  >
>  >  > So as I understand it, from the perspective of the requirements placed
>  >  > on a system if it is to meet the REST architectural principles, using
>  >  > http://myserver.com/invoice <http://myserver.com/invoice> to refer to
>  > a particular resource is no
>  >  > better than using http://myserver.com/akjofjm,
>  > <http://myserver.com/akjofjm,> neither is it any worse.
>  >
>  > Which means that you have to (for REST) assume the latter as there is
>  > nothing to say that http:.//myserver.com/invoice isn't just a random
>  > set of letters that just happen to spell an english word.
>
>  It means that, _in_my_understanding_ of the statements made by REST
>  experts on this list, one is as RESTful as the other. Other
>  considerations may affect your choice between the two of course.

This is a cop-out IMO.  Given that the URI is so important in REST it
would be logical to me that REST URIs should always have a level of
meaning, appending a GUID for a resource would indicate uniqueness,
but having sensible names is just... well sensible.

Saying there is no URI naming standard when the URI is such an
important element is very odd.

>
>  > Either URIs should be well named, or they should be obfuscate, a
>  > half-way "I wish" house is not a good foundation.
>
>  I don't follow your logic here. Are you arguing that because REST does
>  not include any constraint *requiring* meaningful URI names (which as I
>  recall was the misconception that led to this thread) it should
>  therefore *require* them to be obfuscated?

Yes.  By making it optional it in effect requires people to assume the later.

Personally I'm suprised that the answer wasn't "sensible names with a
GUID as the resource identifier".

>                    

Reply via email to