[
Lists Home |
Date Index |
Thread Index
]
Rick Jelliffe wrote:
>My golden retriever Pip does not believe that there can be resources
>that cannot be retrieved. He has done substantial empicial work to
>prove this, and I believe he would respond to an invitation to join the TAG
>enthusiastically, having an enormous understanding of REST.
>
I bet.
>Resources for retrieval can be anonymous, or pointed to (by him
>or me) or named (he only understands the scheme "fetch:" however.)
>
This suggests that your golden retreiver needs nothing more besides the
location (where the resource can be found) and (perhaps) the fetch:
scheme (;-) thus no names are requiered for the function of retreival;
the location is the name because it resolves into one physical unit thus
is unique and proper as a name.
When beyond the stage of retreival being the only function, the location
is just as good as a name since it resolves not to a space but the
actual resource thus is unique. The choices here are either:
* consider the location to be valid for both functions (naming and
resolution) and be prepeared for failing attampts of retreival (404s in
our case) or,
* use the fetch scheme to distinguish between the resources that can be
retreived.
This shows us that all schemes designed to locate something, are as good
for names as they are for locations, when those resolve not to a space
for resources but to an actual resource. Starting to map all this to
URIs, one may concude that an unknown URL can safelly be used as a name
since it's a URI, but must point to a resource (and not wait for a
default resource in that space) to be used as a locator. If it does it's
valid even if that resource does not really exist; everyone can handle
404s, besides that's what 404s where designed for.
So everything is safe if you do not use schemes for retreival if they
where not designed for it and, freely use schemes designed to locate
resources as both names and locators if you are dealing with a
retreivable resource.
Back to your dog now, he has no problem with the fetch: scheme being the
actual command since he only understands physically accesible resources,
whether he is able to retreive them or not. We on the other hand,
distinguish between the schemes and the commands that trigger the
functions these schemes are capable of, since schemes should be used to
denote the avaliable functionality one can use the URI for.
Then again, Pip does not need to worry about that and I think he is in a
better position than I am.
>After playing hide-and-seek recently, he has become convinced of
>my bilocationism:
>
Ok I'll try this one too. So, URL, able of being both a name and a
locator, may be used as a local name for something that is not actually
where it locates it, if the resource space responds by fetching a remote
resource for us.
I wish Pip could share his lights here.
Cheers,
Manos
|
